Description:FIELD OF THE INVENTION
The present invention broadly relates to environmental monitoring. More particularly, the present invention relates to a system and method for assessing water quality through a uniquely designed chatbot interface, that provide real time answers to all water quality-related queries for better understanding of water quality index through different visual representations of water quality parameters. Especially, advanced hybrid machine learning models are deployed to analyse the increase and decrease of various water quality parameters over a certain period of time in order to provide alerts to end users about the water quality.

BACKGROUND OF THE INVENTION
Maintaining water quality is a very huge on-going concern due to the increasing demand for clean and pristine water. Several sectors, including agriculture, urban development, and industries, affect water systems, making it very important to monitor water quality effectively and assess its variation/changes in coming future, so that the precautionary actions or corrective measures can be initiated. Traditionally, water quality assessments have relied on manual data collection and analysis, which can be time-consuming and require skilled workforce. The wide range of water related data are easily procured using various sensors and laboratory testing, but there is no easy way/technique for performing tasks of auditing and presentation of such huge data to predict its environmental impacts in desired formats for environmental researchers/analysts or water board authorities.

However, the recent advancements in artificial intelligence (AI) have opened new avenues for the development of intelligent systems that can help in real-time water quality assessment, data interpretation, and visualization. With these tools, users can access water quality data, request specific analyses, and generate customized visual representations tailored to their requirements.

A reference may be made to CN106198909B that discloses aquaculture water quality prediction technique using a deep learning network having three layers of limited Boltzmann machine (RBM) and one layer of BP neural network.

Another reference may be made to WO2022160682A1 that discloses a water quality monitoring data analysis method for determining occurrence time of pollution by performing correlation analysis on the water quality monitoring data of the reference time period.

Although, the existing AI based water monitoring systems are focused on determining one or more specific components present in the water bodies or at some specific points, however, there are several limitations in terms of computational efficiency, prediction accuracy, results/output delivery, and complexity in configurational setup. Therefore, there is a need of further exploration and investigation to effectively manage and analyse large number of present and past water related data for predicting water characteristics in future environmental conditions, and accordingly deliver the prediction results in user-friendly visual representation formats (such as graphs, text and numerical values) in real-time.

Additionally, the integration of advanced machine learning (ML) can enable predictive modelling, allowing for the early detection of potential issues, such as pollution spikes or infrastructural failures like leaks. This proactive approach to water management helps in efficiently addressing problems before they escalate, thus preventing large-scale environmental damage or waterborne health crises. Incorporating NLP (natural language processing) techniques into water monitoring systems can further enhance user experience by making complex water quality data accessible through natural language queries. Whether users need insights on specific contaminants, historical water quality trends, or recommendations for water treatment, NLP-driven chatbots or platforms may provide a seamless interaction. Moreover, these techniques can generate detailed visualizations, such as time-series graphs or contamination heatmaps, tailored to individual user needs. Further, a hybrid machine learning approach can transform water quality assessment into a more dynamic, accessible, and insightful process, ensuring that decision-makers and the public can act on reliable information quickly.

Therefore, there arises a pressing need for developing an improved approach, system, and method for analysing all changes/fluctuations occurring in water characteristics data and predicting their impact on future water quality and environment, especially providing real-time answers to water quality-related queries in real-time. Moreover, it is desired to develop a technically a method and system for assessing future water quality through advanced machine learning enabled chatbot interface, which includes all the advantages of the conventional/existing techniques/methodologies and overcomes the deficiencies of such techniques/methodologies.

OBJECT OF THE INVENTION
It is an object of the present invention to offer a user-friendly and cost-effective solution for water quality monitoring in various environments.

It is another object of the present invention to develop a chatbot-based water quality assessment tool that can provide real-time answers to water quality-related queries for better understanding of water quality index through different visual representations of water quality parameters.

It is one more object of the present invention to analyse trends of water characteristics changes/fluctuations based on collected data, and generate detailed visualizations (e.g., graphs, charts) that represent water quality over time or across different locations.

It is a further object of the present invention to devise a system and method for assessing future water quality through advanced hybrid machine learning techniques.

SUMMARY OF THE INVENTION
In one aspect, the present invention provides a system for assessing future water quality through advanced machine learning enabled chatbot interface, that can provide answers to all water quality-related queries in real-time. The system comprises a sensor module for measuring characteristics of water samples; a signal conditioner for filtering and converting the measured characteristics into digital data format; a chatbot user interface for answering user queries; and a server communicatively coupled with the signal conditioner and the chatbot user interface. The server equipped with a machine learning trained database associated with water characteristics such as pH, TDS (Total Dissolved Solids), EC (Electrical Conductivity), nitrate, total hardness, calcium, phosphate, and sodium, etc. The chatbot user interface has a query input receiving unit/box and a response output delivering unit/box. The chatbot user interface is embedded with a query processing module configured to: tokenize the query input string into tokens, remove irrelevant tokens, lemmatize remaining tokens, and identify entities and intents behind the query based on the lemmatized tokens. The chatbot user interface is embedded with a sensor data processing module configured to: analyse the digitized data of the measured water characteristics based on the identified entities and intents of the query, and generate response output in visual forms to be displayed in real-time based on the analysed data. The query processing module deploys natural language processing technique for analysing input text data (understanding user queries and generating answers). The sensor data processing module deploys Long Short-Term Memory (LSTM) Networks, seasonal-trend decomposition, moving average, and a time-series linear regression techniques for processing/analysing the water characteristics data and predicting impact on future water quality and environment.

Other aspects, advantages, and salient features of the present invention will become apparent to those skilled in the art from the following detailed description, which delineate the present invention in different embodiments.

BRIEF DESCRIPTION OF DRAWINGS
These and other features, aspects, and advantages of the present invention will become better understood when the following detailed description is read with reference to the accompanying figures.

Fig. 1 is a schematic diagram illustrating key hardware components of the system for assessing water quality, in accordance with an embodiment of the present invention.

Fig. 2 is a block diagram illustrating step-wise backend operations of chatbot user interface, in accordance with an embodiment of the present invention.

Fig. 3 illustrates method steps for assessing water quality, in accordance with an embodiment of the present invention.

Fig. 4 illustrates graphical comparison of various performance metric between the conventional machine learning models and the proposed hybrid machine learning model.

List of reference numerals
100 sensor module
200 signal conditioner
300 server
400 chatbot user interface
Q query
R response

DETAILED DESCRIPTION OF THE INVENTION
Various embodiments described herein are intended only for illustrative purposes and subject to many variations. It is understood that various omissions and substitutions of equivalents are contemplated as circumstances may suggest or render expedient, but are intended to cover the application or implementation without departing from the scope of the present invention. Also, it is to be understood that the phraseology and terminology used herein is for the purpose of description and should not be regarded as limiting.

The use of terms “including,” “comprising,” or “having” and variations thereof herein are meant to encompass the items listed thereafter and equivalents thereof as well as additional items. The terms, “an” and “a” herein do not denote a limitation of quantity, but rather denote the presence of at least one of the referenced items.

In accordance with an embodiment of the present invention, as shown in Fig. 1, the system for assessing water quality is depicted. The system comprises a sensor module (100); a signal conditioner (200); a chatbot user interface (400); and a server (300) communicatively coupled with the signal conditioner (200) and the chatbot user interface (400). The sensor module (100) is configured to measure characteristics of water samples. The signal conditioner (200) is configured to filter and convert the measured characteristics into digital data format. The chatbot user interface (400) has a query input receiving unit (Q) and a response output delivering unit (R). The chatbot user interface (400) employs a query processing module configured to: tokenize the query input string into tokens, remove irrelevant tokens, lemmatize remaining tokens, and identify entities and intents behind the query based on the lemmatized tokens. The chatbot user interface (400) employs a sensor data processing module configured to: analyse the digitized data of the measured water characteristics based on the identified entities and intents of the query, and generate response output in visual forms to be displayed in real-time based on the analysed data. The system is designed to provide answers to all water quality-related queries in real-time.

In accordance with an embodiment of the present invention, the sensor module (100) includes plurality of sensors for reading the water characteristics associated with pH, TDS (Total Dissolved Solids), EC (Electrical Conductivity), nitrate, total hardness, calcium, magnesium, phosphate, and sodium etc. The sensors employ different sensing mechanisms, for example, conductivity sensors detect water by measuring changes in electrical conductivity when water comes into contact with conductive elements.

In accordance with an embodiment of the present invention, the signal conditioner (200) is configured to filter and convert the measured characteristics into digital data format. The signal conditioner (200) includes analogue-to-digital converter The signal captured by the sensors very often contain noise (irrelevant data) and are in analogue format, therefore the noise filtration and the analogue to digital conversion of data are required. The measured water characteristic values are displayed in the interface display.

In accordance with an embodiment of the present invention, the server (300) comprises a memory for storing data and processor executable instructions/codes, and a processor configured to execute the instructions/codes to carry out one or more operational steps of processing user queries, analysing water characteristic data, and generating visual representation as answers to the user queries. The sensor data are stored in CSV (comma-separated values) file format in the memory. The server (300) is further equipped with a machine learning trained database associated with water characteristics.

In accordance with an embodiment of the present invention, the chatbot user interface (400) is configured as a mobile App or web-based application interface where the users input their water related queries in the query input box (Q) and gets correct answers in the response output box (R). The chatbot user interface (400) supports natural language queries and delivers responses either as textual data or visual representations as per the users’ query in real-time.

In accordance with an embodiment of the present invention, the query processing module deploys natural language processing (NLP) technique for analysing input text data (understanding user queries and generating answers).

In accordance with an embodiment of the present invention, the sensor data processing module deploys Long Short-Term Memory (LSTM) Networks, seasonal-trend decomposition technique, moving average technique, and a time-series linear regression technique for processing/analysing the water characteristics data and predicting impact on future water quality and environment. An advanced hybrid model is developed by integrating LSTM, seasonal-trend decomposition, moving average, and time-series linear regression, trained on the database containing water quality parameter values of different regions in Indian territory. With the help of such hybrid model, the system can effectively track water quality trends, detect variations, and provide insights into potential changes, ensuring a proactive approach to environmental management.

In accordance with an embodiment of the present invention, as shown in Fig. 2, the underlying operations of the chatbot user interface (400) is depicted. The underlying (backend) operation includes receiving query input, tokenization, stop word removal, lemmatization, entity and intent recognition, vectorization, query understanding, response generation, and response output delivery.

In accordance with an embodiment of the present invention, the user types queries related to water quality of target regions in the query input box (Q) of the chatbot user interface (400). In tokenization, the input text strings split into smaller individual parts/units, called as tokens. A token can be a word, punctuation mark, or a number which can be processed further in subsequent steps. Let the user input be U, then tokenization divides U into a sequence of tokens as represented in equation 1
U=[token1,token2,…,token n] equation 1

In the stop word removal, the common words that don’t contribute significant meaning to the query are eliminated/discarded. For example, the words like "me", "the", "is", "for", etc., are generally removed as they don’t help in understanding the query. Let T=[token1,token2,…,token n] be the tokens after tokenization, then the stop ward removal produces a revised token set (T^') as indicated in equation 2.
T'=T ∖ {stopwords} equation 2
where ‘\’ represents set difference

The lemmatization helps to reduce different forms of a word to one single form. It converts words into their base form or lemma, thus reducing the dimensionality of the tokenized words. For example, "running" becomes "run", "better" becomes "good". For each word w∈T', its corresponding lemma, lemma(w) are found, then it results in a lemmatized set of tokens (T'') as indication in equation 3.
T^''=[lemma(w1),lemma(w2),…,lemma(wn)] equation 3

The entities (i.e., key words indicating water characteristics or parameters) are identified and the intent behind the query are detected in the lemmatized output. A named entity recognition (NER) model is applied on the lemmatized set of tokens (T^'' ) to detect entities. A probabilistic model like Conditional Random Fields (CRF) is used here. Then, a classification model (such as logistic regression or neural networks) is used on the cleaned tokens to map the query to its intent (I) as indicated in equation 4.
I=argmaxiP(Ii∣T'') equation 4
where P(Ii∣T'') is the probability that the intent is IiI given the tokenized query T''

In vectorization, the cleaned and processed texts are converted into numerical vectors that can be used by machine learning (ML) models. The word2vec model or BERT (Bidirectional Encoder Representations from Transformers) is used to convert each token into a vector based on word embeddings as indicated in equation 5.
Vector(T'')=[v(w1),v(w2),…,v(wn] equation 5
where v(wi) is the embedding vector for the word wi

Based on the identified intent and entities, the system understands what exactly needs to be done based on the input query, and accordingly generates output representations (responses) that are displayed through the interface screen in real-time. Further, the query processing through the chatbot user interface is explained with an example as shown in Table 1.
Table 1
User Input (U): Showing me a scatter plots for the pH and the EC levels
Tokenization (T): [‘Showing’, ‘me’, ‘a’, ‘scatter’, ‘plots’, ‘for’, ‘the’, ‘pH’, ‘and’, ‘the’, ‘EC’, ‘levels’]
Stop-word Removal (T'): ["Showing", "scatter", "plots", "pH", "EC", "levels"]
Lemmatization (T^''): ["Show", "scatter", "plot", "pH", "EC", "level"]
Entity recognition: ["pH", "EC"]
Intent detection: ["Create Scatter Plot"]
Vectorization: word2vec model or BERT is applied
Query understanding: User wants scatter plot using pH and EC parameters of water
Response generation: Scatter plot is generated using pH and EC data
Response output:

In an example, the user collects water samples from various parts of a target region, takes readings of the water samples using the sensors of the system, and the water characteristic data are stored in the server for further processing/analysis. Then, the user asks the chatbot user interface to display water quality index (WQI) of that region. The water quality index is dependent on the following parameters as shown in Table 2.
Table 2
pH: Essential for assessing water acidity.
TDS (Total Dissolved Solids): Affects water taste and salinity.
EC (Electrical Conductivity): Directly linked to TDS and general water quality.
Nitrate (NO3 mg/l): Indicates contamination from fertilizers, important for drinking water.
Total Hardness (TH): Impacts water hardness, critical for both domestic and industrial uses.
Calcium (Ca mg/l) and Magnesium (Mg mg/l): These affect hardness and water usability.
Phosphate (PO4 mg/l): Indicative of pollution and eutrophication potential.
Sodium (Na mg/l): Important for water used in irrigation and human consumption.

Based on how much these parameters contribute to the WQI, every parameter is given a weight, further to scale down the WQI from 0 to 100 every parameter is given a sub-index based on the parameters range according to the Bureau of Indian Standards for water quality, as shown in Table 3.
Table 3
Parameter Value Range Sub-index Description
pH 6.5 ≤ value ≤ 8.5 95 Excellent
pH 6.0 ≤ value < 6.5 or 8.5 < value ≤ 9.0 85 Good
pH 5.5 ≤ value < 6.0 or 9.0 < value ≤ 9.5 60 Average
pH 5.0 ≤ value < 5.5 or 9.5 < value ≤ 10.0 40 Below Average
pH value < 5.0 or value > 10.0 20 Poor
EC (µS/cm) value ≤ 250 95 Excellent
EC (µS/cm) 250 < value ≤ 500 85 Good
EC (µS/cm) 500 < value ≤ 1000 60 Average
EC (µS/cm) 1000 < value ≤ 1500 40 Below Average
EC (µS/cm) value > 1500 20 Poor
TDS (mg/l) value ≤ 500 95 Excellent
TDS (mg/l) 500 < value ≤ 1000 85 Good
TDS (mg/l) 1000 < value ≤ 2000 60 Average
TDS (mg/l) 2000 < value ≤ 3000 40 Below Average
TDS (mg/l) value > 3000 20 Poor
NO3 (mg/l) value ≤ 10 95 Excellent
NO3 (mg/l) 10 < value ≤ 20 85 Good
NO3 (mg/l) 20 < value ≤ 45 60 Average
NO3 (mg/l) 45 < value ≤ 50 40 Below Average
NO3 (mg/l) value > 50 20 Poor
TH as CaCO3 value ≤ 75 95 Excellent
TH as CaCO3 75 < value ≤ 150 85 Good
TH as CaCO3 150 < value ≤ 300 60 Average
TH as CaCO3 300 < value ≤ 600 40 Below Average
TH as CaCO3 value > 600 20 Poor
Ca (mg/l) value ≤ 30 95 Excellent
Ca (mg/l) 30 < value ≤ 75 85 Good
Ca (mg/l) 75 < value ≤ 150 60 Average
Ca (mg/l) 150 < value ≤ 200 40 Below Average
Ca (mg/l) value > 200 20 Poor
Mg (mg/l) value ≤ 10 95 Excellent
Mg (mg/l) 10 < value ≤ 30 85 Good
Mg (mg/l) 30 < value ≤ 100 60 Average
Mg (mg/l) 100 < value ≤ 150 40 Below Average
Mg (mg/l) value > 150 20 Poor
PO4 (mg/l) value ≤ 0.1 95 Excellent
PO4 (mg/l) 0.1 < value ≤ 0.2 85 Good
PO4 (mg/l) 0.2 < value ≤ 0.3 60 Average
PO4 (mg/l) 0.3 < value ≤ 0.5 40 Below Average
PO4 (mg/l) value > 0.5 20 Poor
Na (mg/l) value ≤ 20 95 Excellent
Na (mg/l) 20 < value ≤ 50 85 Good
Na (mg/l) 50 < value ≤ 100 60 Average
Na (mg/l) 100 < value ≤ 150 40 Below Average
Na (mg/l) value > 150 20 Poor

The WQI is calculated using equation 6.
WQI = ∑(_i=1^n)(w_i × SI_i) equation 6
where:
w_i = Weight of the ith parameter
SI_i = Sub-index of the ith parameter
n = Total number of parameters

Based on the WQI values, the system determines the water quality as shown in Table 4.
Table 4
WQI values Indication
91-100 Water is fit for all uses without treatment (Excellent quality)
71-90 Water is clean and safe for most purposes (Good quality)
51-70 Water is moderately polluted, might need some treatment (Average quality)
26-50 Water is polluted and requires substantial treatment (Below average quality)
0-25 Water is heavily polluted and unsuitable for most uses (Poor quality)

In accordance with an embodiment of the present invention, the response is generated using a visual representation module that generates charts, graphs, or other visual outputs based on the user queries. The user can specify the type of visualization, such as Line graphs for time-series data (e.g., pH level over the past month), Bar charts for comparisons between different locations, Heat maps for spatial distribution of water quality indicators, Scatter Plots, Area Plots, and Box Graphs etc.

In accordance with an embodiment of the present invention, for the response generation, the Long Short-Term Memory (LSTM) model is trained in the water characteristics data for detecting long-term dependencies of vector features/values and tracking changes (e.g., increases or decreases) of vector features/values over time in the sensor data. The LSTMs are recurrent neural networks that are highly effective for sequential or time-series data. The LSTM architecture comprises an input layer, middle layers, and output layer. The input layer takes the sequence of historical data (e.g., water quality measurements over days, months, or years). The middle layers include a forget gate, an input gate, and output gate. The forget gate allows the model to determine which information from previous time steps is important and which can be discarded. The input gate decides what new information to add from the current time step. The output gate decides which part of the internal state should be output to predict the trend. These gates help track the rising or falling trends by learning from the historical sequence of data. The output Layer provides predictions about future values or indicates trends (increase/decrease) in the parameters such as pH levels over time, then the LSTM model can learn whether the pH is increasing, decreasing, or remaining stable based on the patterns in the historical data.

In accordance with an embodiment of the present invention, for the response generation, the seasonal-trend decomposition technique is used for separating seasonal variations from usual trends found in the water characteristics data as sensed by the sensors. The Seasonal-Trend Decomposition is a statistical method that separates trend, seasonal, and residual components from a time-series dataset. By isolating the trend component, it can track whether a parameter is increasing or decreasing over time. The trend components (regular data fluctuation) represent underlying movement in the data (e.g., changes in water quality due to regular increasing or decreasing of values over time). The Seasonal Components (periodical data fluctuation) represent any repeating patterns (e.g., periodic fluctuations in water quality due to seasonal changes). The Residual Components represent noise or irregularities left after removing the trend and seasonal components. The decomposition formula is given in equation 7.
Y(t)=T(t)+S(t)+R(t) equation 7
where,
Y(t): observed value at time t,
T(t): trend component,
S(t): seasonal component,
R(t): residual component.

In an example, the user wants to see trends of TDS values over several years. The system separates the seasonal variation (e.g., higher during the summer due to evaporation) from the underlying usual trends (e.g., gradual increase in TDS levels year-over-year) using the seasonal-trend decomposition.

In accordance with an embodiment of the present invention, for the response generation, the moving average technique is used for smoothening out short-term variations and highlighting long-term trends found in the water characteristics data as sensed by the sensors. A moving average is calculated over a sliding window of time (t) using equation 8.
Moving Average(t)=(x(t)+x(t-1)+⋯+x(t-n))/n equation 8
where
x(t) is the value at time t,
n is the size of the sliding window

In an example, the user can see the moving average of pH levels over a 30-day window to smooth out daily variations and observe if the pH is gradually increasing or decreasing over months or years.

In accordance with an embodiment of the present invention, for the response generation, the time-series linear regression technique is used for determining changes in the water characteristic values over time. The Linear regression can model the relationship between time (as an independent variable) and a water quality parameter (as a dependent variable) to determine whether the parameter is increasing or decreasing. A linear trend in a time-series dataset is calculated using equation 9.
y=mx+ c equation 9
where
y= water quality parameter (e.g., pH, TDS),
x = time (in days, months, or years),
m = slope (rate of change, indicating whether the parameter is increasing or decreasing),
c = intercept.

In an example, the system can fit a linear regression model to the pH levels over time and look at the slope mmm. If m>0m > 0m>0, the parameter is considered as increasing over time; if m<0m < 0m<0, it is considered as decreasing.

In accordance with an embodiment of the present invention as shown in Fig. 3, the method for assessing water quality is depicted. The method employs a sensor module (100); a signal conditioner (200); a chatbot user interface (400); and a server (300). The method comprises steps of: measuring (S1) the water sample characteristics followed by filtration and conversion into digital data format; receiving (S2) the query inputs from the users; tokenizing (S3) each string of the query inputs into tokens, followed by removal of irrelevant tokens, lemmatization of remaining tokens, and identification of entities and intents behind the query based on the lemmatized tokens; analysing (S4) the digitized data of the measured water characteristics based on the identified entities and intents of the query, followed by response output generation in visual forms; displaying (S5) the response output in real-time.

In accordance with an exemplary embodiment of the present invention, the sensor data analysis step (S4) comprises: detecting long-term dependencies and tracking changes over time found in the water characteristics data through the Long Short-Term Memory (LSTM) Networks; separating seasonal variations from usual/regular trends found in the water characteristics data through seasonal-trend decomposition technique; smoothening out short-term variations and highlighting long-term trends found in the water characteristics data through moving average technique; and determining changes in the water characteristic values over time through time-series linear regression technique.

Referring to Fig. 4, the proposed hybrid model is compared with the conventional (primitive/baseline) models to check their response output differences in terms of three key metrics such as accuracy, FI score, and computation time.

It is observed that the proposed hybrid model achieves an accuracy of 90% and an F1 score of 87%, outperforming the individual primitive LSTM, seasonal-trend decomposition (STL), and Random Forest models. This indicates that the proposed model is more reliable in predicting water quality trends and parameters such as pH, EC, and TDS etc.

It is also observed that the proposed hybrid model has a significantly lower computation time compared to LSTM and seasonal-trend (STL), making it more efficient for real-time applications. It only takes 1.2 seconds to process data, enabling faster insights and decision-making, which is crucial for water quality monitoring.

Further, the present invention the following technical advantages including but not limited to:
While the proposed hybrid model is designed in a unique and optimized way to enhance performance. For example, instead of just using an LSTM for time-series prediction, the model integrates it with seasonal-trend decomposition and linear regression to not only detect trends but also capture seasonality and fluctuations more accurately, which traditional models fail to capture effectively. The proposed hybrid approach allows for more precise water quality predictions compared to using these models independently.
The proposed hybrid model is designed specifically for real-time water quality monitoring and integrates machine learning models in an environment that is optimized for low-latency decision-making. This is achieved through efficient data processing pipelines and the real-time response of the chatbot interface. Traditional models, when used alone, are often computationally expensive and not suitable for real-time insights. By tailoring these models for real-time applications, the present invention significantly reduces computational overhead without sacrificing accuracy.
The proposed hybrid model is highly specialized for water quality parameters such as pH, EC, and TDS, etc. while the traditional models may be more general-purpose. By fine-tuning the models to suit specific water quality trends and factors, the present invention achieves better performance.
In addition to superior model integration, the present invention provides a unique user experience with natural language processing (NLP) and customized visualizations. This allows end users, even those without technical expertise, to access water quality insights seamlessly. Standard models like Random Forest or standalone LSTMs do not offer such an intuitive interface for real-time data queries and visualizations, which gives the proposed model an edge in practical applications.
The proposed model employs multiple ML architectures for trend detection in combination, which leads to more robust predictions by balancing short-term and long-term dependencies in the data. This is a significant improvement over using any one model in isolation. By fusing these methods, the present invention is able to better handle both sudden changes and gradual trends in the water quality data.

The foregoing descriptions of exemplary embodiments of the present invention have been presented for purposes of illustration and description. They are not intended to be exhaustive or to limit the invention to the precise forms disclosed, and obviously many modifications and variations are possible in light of the above teaching. The exemplary embodiment was chosen and described in order to best explain the principles of the invention and its practical application, to thereby enable the persons skilled in the art to best utilize the invention and various embodiments with various modifications as are suited to the particular use contemplated. It is understood that various omissions, substitutions of equivalents are contemplated as circumstance may suggest or render expedient, but is intended to cover the application or implementation without departing from the scope of the claims of the present invention. , Claims:We claim:

1. A system for assessing water quality, comprises:
a sensor module (100) configured to measure characteristics of water samples;
a signal conditioner (200) configured to filter and convert the measured characteristics into digital data format;
a chatbot user interface (400) having a query input receiving unit (Q) and a response output delivering unit (R);
a server (300) equipped with a machine learning trained database associated with water characteristics, and communicatively coupled with the signal conditioner (200) and the chatbot user interface (400);
wherein the chatbot user interface (400) is embedded with:
a query processing module configured to: tokenize the query input string into tokens, remove irrelevant tokens, lemmatize remaining tokens, and identify entities and intents behind the query based on the lemmatized tokens;
a sensor data processing module configured to: analyse the digitized data of the measured water characteristics based on the identified entities and intents of the query, and generate response output in visual forms to be displayed in real-time based on the analysed data.

2. The system as claimed in claim 1, wherein the sensor module (100) includes plurality of sensors for reading the water characteristics associated with pH, TDS (Total Dissolved Solids), EC (Electrical Conductivity), Total Hardness, nitrate, calcium, magnesium, phosphate and sodium.

3. The system as claimed in claim 1, wherein the signal conditioner (200) includes analogue-to-digital converter.

4. The system as claimed in claim 1, wherein the query processing module deploys natural language processing technique for analysing input text data.

5. The system as claimed in claim 1, wherein the sensor data processing module deploys Long Short-Term Memory (LSTM) Networks for detecting long-term dependencies and tracking changes over time found in the water characteristics data.

6. The system as claimed in claim 1, wherein the sensor data processing module deploys a seasonal-trend decomposition technique for separating seasonal variations from regular trends found in the water characteristics data.

7. The system as claimed in claim 1, wherein the sensor data processing module deploys a moving average technique for smoothening out short-term variations and highlighting long-term trends found in the water characteristics data.

8. The system as claimed in claim 1, wherein the sensor data processing module deploys a time-series linear regression technique for determining changes in the water characteristic values over time.

9. A method for assessing water quality, comprises steps of:
measuring (S1), by a sensor module, water sample characteristics followed by filtration and conversion into digital data format, wherein the digital data are stored in a water characteristics database in a server;
receiving (S2), by a chatbot user interface, query inputs from users;
tokenizing (S3), by a query processing module, each string of the query inputs into tokens, followed by removal of irrelevant tokens, lemmatization of remaining tokens, and identification of entities and intents behind the query based on the lemmatized tokens;
analysing (S4), by a sensor data processing module, the digitized data of the measured water characteristics based on the identified entities and intents of the query, followed by response output generation in visual forms; and
displaying (S5), by the chatbot user interface, the response output in real-time.

10. The system as claimed in claim 9, wherein the sensor data analysis step (S4) comprises:
detecting long-term dependencies and tracking changes over time found in the water characteristics data through Long Short-Term Memory (LSTM) Networks;
separating seasonal variations from regular trends found in the water characteristics data through seasonal-trend decomposition technique;
smoothening out short-term variations and highlighting long-term trends found in the water characteristics data through moving average technique; and
determining changes in the water characteristic values over time through time-series linear regression technique.