The present invention relates to the field of language processing. More specifically, present invention relates to a Punjabi grammar checking system wherein the tagger as well the whole system is functionalized through multi-layer neural networks for providing real-time error detection and correction with a consistently well accuracy on Punjabi corpus.
BACKGROUND OF THE INVENTION
Background description includes information that may be useful in understanding the present invention. It is not an admission that any of the information provided herein is prior art or relevant to the presently claimed invention, or that any publication specifically or implicitly referenced is prior art.
Detecting grammatical errors in a text has always remained a major concern for society. Due to unavoidable human mistakes, although we try our best to write/type text on a page/word file, still some grammatical errors always remain. One more reason for the same may be less knowledge of a user in a particular language in which he/she may desire to write.
Multiple reasons have led to the invention of multitude of spelling and grammar error detection and correction systems. Such systems assist the user to digitally upload their work file/scanned file of text, and provide suggestions on grammar errors and suggestions on the same. Such grammar error detection and correction systems have been designed in many different languages.

US9037967B1 discloses one such Arabic spelling error detection and correction method for identifying real-word spelling errors. The method uses a corpus of Arabic text alongside n-gram statistical techniques to detect erroneous words within the text. After identifying the erroneous word the method uses a dictionary formed from the corpus of Arabic text to retrieve candidate correction word to replace the erroneous word. Using the n-gram statistical model candidate correction words are generated and ranked in order of highest probable correction for the word. The generated and ranked correction words are assessed and the best correction word is selected. A final assessment of the correction is conducted and if the result is positive then erroneous word is replaced with the highest statistical correction.
KR20160084670A discloses a sentence error correcting apparatus and a method for a conversation based language learning system, which can exactly detect an error in an inputted sentence and correct the error using information on the context of a sentence or the like related to the inputted sentence as well as information on the context of the inputted sentence, in the case of detecting an error in the inputted sentence. According to an embodiment of the present invention, the sentence error correcting apparatus comprises a detection unit that analyzes the context of an inputted present conversational sentence, and the context of a conversational sentence pre-stored before the present conversational sentence, and which detects an error in the present conversational sentence using information on the analyzed contexts and a correction unit which corrects the error in the present conversational sentence using information on the error detected by the detection unit.
Several works have been performed so far to efficiently detect and correct grammatical errors in multiple languages, but all such systems and methods the perceptron tagger involved therein depends only on the part of speech sequence information and not on the vocabulary, parts of speech, meanings and contextual co-occurrence relations of surrounding works which requires a proper language

lexicon as an input and efficient preparation of relevant tags. Moreover, real-time detection and correction of errors has not possible in works done till now in the context of grammar checking systems.
Hence, our disclosure is directed towards a grammar checking system having an ameliorated perceptron tagger as well as the ability to detect errors and suggest corrections in real-time for the Punjabi language, and compatible enough to design a modifiable fit for other Indian languages as well.
OBJECTS OF THE INVENTION
The principal object of the present invention is to overcome the disadvantages of the prior art.
An object of the present invention is to provide a Punjabi grammar checking system ameliorated through neural networks.
Another object of the present invention is to provide a Punjabi grammar checking system having an ameliorated perceptron tagger trained through error backpropagation neural network protocol that provides consistently well accuracy.
Another object of the present invention is to provide a Punjabi grammar checking system that encompasses the complete corpus (all the words, phrases, morphemes) of the Punjabi language.
Another object of the present invention is to provide a Punjabi grammar checking system having a perceptron tagger configured to efficiently handle unknown and ambiguous words in a limitedly resourced Punjabi language.
Another object of the present invention is to provide a Punjabi grammar checking system that can provide grammatical correction suggestions as well for

the detected errors.
Another object of the present invention is to provide a Punjabi grammar checking system that works equally well for all the Indian Languages in terms of error detection and correction.
Yet another object of the present invention is to provide a Punjabi grammar checking system that provides real-time error detection as well as correction suggestions by using a multi-layer artificial neural network protocol.
The foregoing and other objects, features, and advantages of the present invention will become readily apparent upon further review of the following detailed description of the preferred embodiment as illustrated in the accompanying drawings.
SUMMARY OF THE INVENTION
The present invention relates a Punjabi grammar checker system having a deeply learned perceptron tagger and error detection module for providing consistent, accurate and real-time detection and correction suggestions for grammatical errors in Punjabi Corpus.
According to an embodiment of present subject matter, the Punjabi grammar checking system ameliorated through neural networks comprises an acquisition unit for receiving atleast one input corpus, a sentence refiner configured to tokenize the input corpus into a set of input morphemes, a sentence analyzer arranged next to the sentence refiner for returning plurality of tags encompassing the set of input morphemes, a perceptron tagger configured to assign the tags to each of the set of input morphemes using error back propagation neural network protocol, thereby formulating a set of tagged morphemes, wherein the perceptron tagger comprises of a training unit and a testing unit for regimenting the tagger through the protocol

by referring a Punjabi lexicon, wherein learnings obtained from the training unit are further checked for variations from the input corpus via a neural comparator, a phrase builder configured to develop an output corpus by examining of degree of closeness of the set of tagged morphemes by parsing thereof using a rule based phrase chunking technique, an error detection module configured to check grammatical errors in the output corpus by applying multi-layer artificial neural network protocol thereon and a user-operable discrepancy reporter to provide a detailed error information and correction suggestions based on the output received through the error detection module.
According to another embodiment of present subject matter, the refiner further filters the input corpus including but not limited to fixed sentences, marking phrases, unreadable sentences and sentences from a language other than that in context. According to another embodiment of present subject matter, the sentence analyzer uses a Punjabi lexicon for returning the plurality of tags.
According to another embodiment of present subject matter, the tags include but not limited to morpheme class, noun's adjective, noun's gender, noun's number, noun's case, pronoun's person, verb's gender, verb's number, verb's person, verb's tense, verb's phrase, verb's transitivity, adverb, particle, cardinal, adjective, interjection and coordinate conjunct. According to another embodiment of present subject matter, the perceptron tagger adjusts weights of direct links between the units of neighboring layers to assign the tags to each of the set of input morphemes.
According to another embodiment of present subject matter, learnings obtained from the training unit include but not restricted to morpheme, morpheme index, frequency of occurrence of a morpheme and tag assigned to a morpheme. According to another embodiment of present subject matter, the perceptron tagger assigns the tags to a collection of known morphemes in the set of input morphemes, based on the learnings obtained from the training unit. According to another

embodiment of present subject matter, the perceptron tagger predicts tags with highest value for a collection of unknown and ambiguous morphemes.
According to another embodiment of present subject matter, the multi-layer artificial neural network protocol is trained on a test corpus, thereby configuring the multi-layer artificial neural network protocol to check for grammatical errors in the output corpus. According to yet another embodiment of present subject matter, the discrepancy reporter is chosen from but not limited to tab, laptop, computer, mobile, register and microphone.
While the invention has been described and shown with particular reference to the preferred embodiment, it will be apparent that variations might be possible that would fall within the scope of the present invention.
BRIEF DESCRIPTION OF THE DRAWINGS
The accompanying drawings are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification. The drawings illustrate exemplary embodiments of the present disclosure and, together with the description, serve to explain the principles of the present disclosure.
In the figures, similar components and/or features may have the same reference label. Further various components of the same type may be distinguished by following the reference label with a second label that distinguishes among the similar components. If only the first reference label is used in the specification, the description is applicable to any of the similar components having the same reference label irrespective of the second reference label.
Figure 1 illustrates a block diagram of a Punjabi grammar checking system ameliorated through neural networks, according to an embodiment.

Figure 2 illustrates a schematic diagram of a perceptron tagger configured to assign tags to each member of set of input morphemes using error backpropagation neural network protocol, according to an embodiment.
Figure 3 illustrates a flow chart of a method followed by the Punjabi grammar checking system ameliorated through neural networks, according to an embodiment.
Figure 4 shows user-operable screen of an exemplary discrepancy reporter to provide detailed error information and correction suggestions, according to an embodiment.
DETAILED DESCRIPTION OF THE INVENTION
As used in the description herein and throughout the claims that follow, the meaning of "a," "an," and "the" includes plural reference unless the context clearly dictates otherwise. Also, as used in the description herein, the meaning of "in" includes "in" and "on" unless the context clearly dictates otherwise.
If the specification states a component or feature "may", "can", "could", or "might" be included or have a characteristic, that particular component or feature is not required to be included or have the characteristic.
Exemplary embodiments will now be described more fully hereinafter with reference to the accompanying drawings, in which exemplary embodiments are shown. This disclosure may however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. These embodiments are provided so that this disclosure will be thorough and complete and will fully convey the scope of the disclosure to those of ordinary skill in the art. Moreover, all statements herein reciting embodiments of the disclosure, as well as

specific examples thereof, are intended to encompass both structural and functional equivalents thereof. Additionally, it is intended that such equivalents include both currently known equivalents as well as equivalents developed in the future (i.e., any elements developed that perform the same function, regardless of structure).
Various terms as used herein are shown below. To the extent a term used in a claim is not defined below, it should be given the broadest definition persons in the pertinent art have given that term as reflected in printed publications and issued patents at the time of filing.
In some embodiments, the numerical parameters set forth in the written description and attached claims are approximations that can vary depending upon the desired properties sought to be obtained by a particular embodiment. In some embodiments, the numerical parameters should be construed in light of the number of reported significant digits and by applying ordinary rounding techniques. Notwithstanding that the numerical ranges and parameters setting forth the broad scope of some embodiments of the invention are approximations, the numerical values set forth in the specific examples are reported as precisely as practicable. The numerical values presented in some embodiments of the invention may contain certain errors necessarily resulting from the standard deviation found in their respective testing measurements.
The present invention relates to a grammar checking system ameliorated through neural networks for efficiently detecting errors in real-time in Punjabi language and subsequently proving correction suggestions as well in a consistently well and accurate manner.
Referring to Figure 1, a block diagram of the Punjabi grammar checking system ameliorated through neural networks is presented, according to an embodiment present disclosure. The grammar checking system comprises of an acquisition unit, a sentence refiner, a sentence analyzer, a perceptron tagger, a

phrase builder, an error detection module and a discrepancy reporter.
The acquisition unit embodies the function of receiving atleast one input corpus. The input corpus referred to herein is the collection of words and/or phrases required to be tested for grammar. A user may provide the input corpus in the acquisition unit as an already prepared document or a text written in real-time can also be word-by-word/phrase-by-phrase continually provided to the acquisition unit.
The sentence refiner can be configured to break the so received input corpus into a set of input morphemes. The set of input morphemes is a collection of smallest meaningful units in the input corpus. As an instance, the word "unbreakable" is made up of three morphemes named as "un", "break", and "able". Similarly, a set of all the morphemes possible in the inputted corpus can be collected using the sentence refiner. Furthermore, the refiner filters the input corpus for including but not limited to fixed sentences, marking phrases, unreadable sentences and sentences from a language other than that in context.
As a next step, the sentence analyzer can be arranged next to the sentence refiner for returning plurality of tags encompassing the set of input morphemes prepared by the sentence refiner. Herein, the sentence analyzer applies morphological analysis on the received input corpus and uses the Punjabi lexicon provided as a standard reference thereto. The Punjabi words can be categorized based on atleast 22 types of tags, preferably including noun, pronoun, adjective, cardinals, verbs, adverbs, ordinals, postposition, conjunction, interjection, and likewise depending on the word under invigilation.
Therefore, the tags include but not limited to morpheme class, noun's adjective, noun's gender, noun's number, noun's case, pronoun (personal, reflexive, demonstrative, relative, interrogative and indefinite), verb's gender, verb's number, verb's person, verb's tense, verb's phrase, verb's transitivity,

adverb, particle, cardinal, adjective (inflected and uninfected), interjection and coordinate conjunct.
Referring to Figure 2, a schematic diagram of a perceptron tagger configured to assign tags to each member of the set of input morphemes using error backpropagation neural network protocol is reported, according to an embodiment of present disclosure. Once the tags are prepared, the perceptron tagger assigns the tags to each of the input morphemes using error backpropagation neural network protocol, thereby formulating a set of tagged morphemes.
As indicated in figure 2, the perceptron tagger comprises of a training unit, a testing unit, a Punjabi lexicon, a neural network process and a comparator. The training unit and the testing unit are functionalized to process the Punjabi lexicon, thus building up a learned neural network-based tagger. The deep learnings obtained by the perceptron tagger using error backpropagation neural network protocol in the training unit can be further applied for checking variations from the input corpus via a neural comparator.
Herein, the perceptron tagger adjusts weights of direct links between the units of neighboring layers to assign the tags to each of the set of input morphemes. Moreover, the learnings obtained from the training unit include but not restricted to morpheme, morpheme index, frequency of occurrence of a morpheme and tag assigned to a morpheme.
Notably, the perceptron tagger adjusts weights of direct links between the units of neighboring layers in the protocol (such as input layer, hidden layer and output layer and other sublayers involved thereinto) to assign the tags to each of the set of input morphemes.
While the tags are being assigned to each member of the set of morphemes, the words usually fall under the categories of known morphemes, unknown

morphemes and ambiguous morphemes. Known morphemes are the trained ones because they are available in the Punjabi lexicon referred for the present invention, unknown morphemes are the ones not trained by the neural network and the ambiguous tags are those which have been assigned more than two tags. Herein, the perceptron tagger predicts tags with the highest value for a collection of unknown and ambiguous morphemes.
Upon comparison of the present invention's perceptron tagger with the Support Vector Machine (SVM) based tagger and Hidden Markov Model (HMM) based tagger, the accuracy can be found to be more enhanced, having a value of 89.99%, as presented in table 1 below:
TABLE 1

CORPUS SIZE PROTOCOL UNDER TEST ACCURACY
26,500 SVM 85.48%
26,500 HMM 81.72%
26,700 error backpropagation neural network protocol 89.99%
Subsequently, the phrase builder develops an output corpus by examining the degree of closeness of the set of tagged morphemes. The output corpus is attained by building up phrase combinations based on the tags allotted to the morphemes through parsing. The method used to accomplish this task includes rule-based phrase chunking technique which works based on certain rules to lay the basis on which the combining of the phrases is done.
As an instance, lack of agreement in words of a particular phrase is

considered as a criterion of not chunking such a phrase. Likewise, in order to maintain accountability of misplaced words, commas are used such that if words of a particular phrase are separated through commas, then they cannot be taken as a single phrase. However, such consideration may lead to potential error as well if the placement of comma may be correct and incorrectly assumed to be a misplaced word.
Practically, the real-time grammar analysis has been made possible through the present invention by the incorporation of an error detection module configured to check grammatical errors in the output corpus by applying a multi-layer neural network protocol thereon. Through the use of a multi-perceptron artificial neural network for error detection and correction, though the training as well as processing time to analyze the results gets increased, but the results so obtained are computationally accurate, predictable and allows real-time assessment. This is because of the human-machine training culture applied through deep learning modules available for error detection training.
Finally, the discrepancy reporter is connected to the error detection module, such that the reporter receives the analysis done by the detection module and presents it in a suitably readable form on a user-operable screen. The discrepancy reporter is chosen from but not limited to tab, laptop, computer, mobile, register and microphone.
The discrepancy reporter can comprise of plurality of buttons to provide various captions desired by a user while viewing the error detection results and the related suggestions like an error message tab, back tab, portion to show the inputted phrase, the incorrect words in the inputted phrase and the subsequent correction suggestions provided to the user.
As indicated in Figure 3, a flow chart of a method followed by a Punjabi grammar checking system ameliorated through neural networks is shown,

according to an embodiment of present disclosure. The method comprises the steps of receiving atleast one input corpus through the acquisition unit, tokenize the input corpus into a set of input morphemes via the sentence refiner, returning plurality of tags encompassing the set of input morphemes with the help of sentence analyzer, using perceptron tagger to assign the tags to each of the set of input morphemes using error backpropagation neural network protocol, thereby formulating a set of tagged morphemes.
Herein, the method followed by perceptron tagger includes the step of regimenting the tagger through the application of error backpropagation neural network protocol in the training unit and testing unit by referring a Punjabi lexicon and thereafter checking for variations from the input corpus via a neural comparator depending on the learnings obtained from the training unit.
In furtherance to this, the method follows the steps of examining the degree of closeness of the set of tagged morphemes by parsing thereof using a rule-based phrase chunking technique in the phrase builder, checking grammatical errors in the output corpus obtained in the error detection module by applying multi-layer artificial neural network protocol thereon, and finally utilizing a user-operable discrepancy reporter to provide a detailed error information and correction suggestions based on the output received through said error detection module.
Referring to Figure 4, the user-operable screen of an exemplary discrepancy reporter is shown to provide detailed error information and correction suggestions, according to an embodiment of present invention. As illustrated in the figure, the disclosed Punjabi grammar checker's discrepancy reporter is user-operable and includes tabs including but not limited to an error message, reason of error screen having the text under study and the related grammatical error with the suggested correction, and lastly a back tab. Overall, everything required by a user while pursuing grammatical error analysis of an input corpus can be incorporated thereon.

The final accuracy of the overall system using the multi-layer artificial neural network has been found to be highly increased having value of 85.67%, which is much higher than that obtained from the systems not using multi-layer artificial neural network for grammatical error detection and correction. However, the processing time to get the output of analyzed errors and suggestions gets increased to a reasonably acceptable amount.
It should be apparent to those skilled in the art that many more modifications besides those already described are possible without departing from the inventive concepts herein. The inventive subject matter, therefore, is not to be restricted except in the spirit of the appended claims. Moreover, in interpreting both the specification and the claims, all terms should be interpreted in the broadest possible manner consistent with the context. In particular, the terms "includes" and "including" should be interpreted as referring to elements, components, or steps in a non-exclusive manner, indicating that the referenced elements, components, or steps may be present, or utilized, or combined with other elements, components, or steps that are not expressly referenced. Where the specification claims refers to at least one of something selected from the group consisting of A, B, C ... .and N, the text should be interpreted as requiring only one element from the group, not A plus N, or B plus N, etc. The foregoing description of the specific embodiments will so fully reveal the general nature of the embodiments herein that others can, by applying current knowledge, readily modify and/or adapt for various applications such specific embodiments without departing from the generic concept, and, therefore, such adaptations and modifications should and are intended to be comprehended within the meaning and range of equivalents of the disclosed embodiments. It is to be understood that the phraseology or terminology employed herein is for the purpose of description and not of limitation. Therefore, while the embodiments herein have been described in terms of preferred embodiments, those skilled in the art will recognize that the embodiments herein can be practiced with modification within the spirit and scope of the appended claims.

While embodiments of the present disclosure have been illustrated and described, it will be clear that the disclosure is not limited to these embodiments only. Numerous modifications, changes, variations, substitutions, and equivalents will be apparent to those skilled in the art, without departing from the spirit and scope of the disclosure, as described in the claims.
ADVANTAGES OF THE INVENTION
The present invention provides a Punjabi grammar checking system ameliorated through neural networks.
The present invention provides a Punjabi grammar checking system having an ameliorated perceptron tagger trained through error backpropagation neural network protocol that provides consistently well accuracy.
The present invention provides a Punjabi grammar checking system that encompasses the complete corpus (all the words, phrases, morphemes) of the Punjabi language.
The present invention provides a Punjabi grammar checking system that can provide grammatical correction suggestions as well for the detected errors.
The present invention provides a Punjabi grammar checking system that works equally well for all the Indian Languages in terms of error detection and correction.
The present invention provides a Punjabi grammar checking system that provides real-time error detection as well as correction suggestions by using multi-layer artificial neural network protocol.
The present invention provides a Punjabi grammar checking system having

perceptron tagger configured to efficiently handle unknown and ambiguous words in a limitedly resourced Punjabi language.

I/We Claim:
1) A Punjabi grammar checking system ameliorated through neural networks,
comprising:
an acquisition unit for receiving atleast one input corpus; a sentence refiner configured to tokenize said input corpus into a set of input morphemes;
a sentence analyzer arranged next to said sentence refiner for returning plurality of tags encompassing said set of input morphemes;
a perceptron tagger configured to assign said tags to each of said set of input morphemes using error backpropagation neural network protocol, thereby formulating a set of tagged morphemes;
wherein said perceptron tagger comprises of a training unit and a testing unit for regimenting said tagger through said protocol by referring a Punjabi lexicon;
wherein learnings obtained from said training unit are further checked for variations from said input corpus via a neural comparator; a phrase builder configured to develop an output corpus by examining of degree of closeness of said set of tagged morphemes by parsing thereof using a rule-based phrase chunking technique;
an error detection module configured to check grammatical errors in said output corpus by applying multi-layer artificial neural network protocol thereon; and
a user-operable discrepancy reporter to provide detailed error information and correction suggestions based on the output received through said error detection module.
2) The system, as claimed in claim 1, wherein said refiner further filters said input
corpus for including but not limited to fixed sentences, marking phrases, unreadable
sentences and sentences from a language other than that in context.

3) The system, as claimed in claim 1, wherein said sentence analyzer morphological analysis and Punjabi lexicon for returning said plurality of tags.
4) The system, as claimed in claim 1, wherein said tags include but not limited to morpheme class, noun's adjective, noun's gender, noun's number, noun's case, pronoun (personal, reflexive, demonstrative, relative, interrogative and indefinite), verb's gender, verb's number, verb's person, verb's tense, verb's phrase, verb's transitivity, adverb, particle, cardinal, adjective (inflected and uninflected), interjection and coordinate conjunct.
5) The system, as claimed in claim 1, wherein said perceptron tagger adjusts weights of direct links between the units of neighboring layers to assign said tags to each of said set of input morphemes.
6) The system, as claimed in claim 1 or 3, wherein learnings obtained from said training unit include but not restricted to morpheme, morpheme index, frequency of occurrence of a morpheme and tag assigned to a morpheme.
7) The system, as claimed in claim 1, wherein said the perceptron tagger adjusts weights of direct links between the units of neighboring layers to assign said tags to each of said set of input morphemes
8) The system, as claimed in claim 1 or 7, wherein said perceptron tagger predicts tags with the highest value for a collection of unknown and ambiguous morphemes.
9) The system, as claimed in claim 1, wherein said multi-layer artificial neural network protocol is trained on a test corpus, thereby configuring said multi-layer artificial neural network protocol to check for grammatical errors in said output corpus.

10) The system, as claimed in claim 1, wherein said discrepancy reporter is chosen from but not limited to tab, laptop, computer, mobile, register and microphone.