23
Guided By, Mrs. Gauri M. Dhopavkar Presented By, Ritikesh Bhaskarwar Vimal Shah Ashwin Borkar Shashil Pohankar

Presentation1

Embed Size (px)

DESCRIPTION

Natural Language Processing, , , , , ,Marathi POS tagger

Citation preview

Page 1: Presentation1

Guided By,Mrs. Gauri M.

Dhopavkar Presented By,Ritikesh Bhaskarwar Vimal ShahAshwin Borkar Shashil Pohankar

Page 2: Presentation1

Department of Computer Technology YESHWANTRAO CHAVAN COLLEGE OF

ENGINEERING, Nagpur (An Autonomous Institution Affiliated to Rashtrasant Tukadoji Maharaj

Nagpur University)

Page 3: Presentation1

Natural language processing

Natural language processing (NLP) is a field of computer science, artificial intelligence, and linguistics concerned with the interactions between computers and human (natural) languages.

Natural Language Processing (NLP) is the computerized approach to analysing text that is based on both a set of theories and a set of technologies

Page 4: Presentation1
Page 5: Presentation1

POS Tagging :

Part-of-Speech (POS) tagging is the process of assigning a part-of-speech like noun, verb, pronoun or other lexical class marker to each word in a sentence.

After POS tags are identified, the next step is chunking, which involves dividing sentences into non-overlapping non-recursive phrases.

Page 6: Presentation1

ते� फू� ल खू�प सुगं�धी� आहे�

Marathi POS Tagger

ते�-unidentifiedफू� ल-nounखू�प-adjectiveसुगं�धी�-adjectiveआहे�-verb

THE POS TAGGING EXAMPLE

Page 7: Presentation1

Need of Marathi POS Tagging :

Lack of significant tools for Indian languages

Dependence of other NLP activities on POS tagging

Failure of existing techniques on Indian Languages

Page 8: Presentation1

Overview of POS tagging

Click icon to add picture

Page 9: Presentation1

Methods for POS Tagging

1.Rule Based 2.Stochastic

The rule based POS tagging models apply a set of hand written rules and use contextual information to assign POS tags to words.

A stochastic approach includes frequency, probability or statistics. The simplest stochastic approach finds out the most frequently used tag for a specific word in the annotated training data and uses this information to tag that word in the unannotated text.

Page 10: Presentation1

Methods for POS Tagging(cntd.)

3. Hiden Markov Model 4. Maximum Entropy Model

The HMM model trains on annotated corpora to find out the transition and emission probabilities

The Maximum Entropy Model (MEM) is based on the principle of Maximum Entropy, which states that when choosing between a number of different probabilistic models for a set of data, the most valid model is the one which makes fewest arbitrary assumptions about the nature of the data

Page 11: Presentation1

Architecture and Design :Marathi sentence is taken as

input , then the tokens are created followed by tagging and finding ambiguity.

TOKENIZING TAGGING FINDING AMBIGUOUS WORDS

FINDING PROBABILITY

ASSIGN TAGS ACCORDING TO PROBABILITY

VIEW THE RESULT

INPUT

Page 12: Presentation1

Detail of Identified Module :Tokenizer : This module is used to

get the tokens of the input sentence. Also, calls the other modules when required.

Tagging : These modules is used for assigning certain tags to tokens and also search for ambiguous words and also find their types and assign some special symbols to them.

Page 13: Presentation1

Details of identified modules (cntd.)Root word : This module is used for

finding the root word of each token finding it from the Marathi wordnet.

Probability : This module calculates the probability and accordingly assigns the tag, according to the higher probability of word.

• Showing the results : This module shows the result. The words are shown with tags.

Page 14: Presentation1

Experimentation and Results :

1.

•1000: If first bit is 1, then we assign a tag as a noun to the particular word.

•1100: In this case, the word can be used as both unidentified.

2.

•0100: If second bit is 1, then we assign a tag as an adjective to the particular word.

•0110: In this case, the word can be used as other words.

3.

•0010: If third bit is 1, then we assign a tag as an adverb to the particular word.

•0001: If fourth bit is 1, then we assign a tag as a verb to the particular word.

Page 15: Presentation1

Advantages :A POS tagger can be seen as a first-step

towards tightening the integration between speech recognition and natural language processing.

A POS tagger in the language model aids in the identification of boundary tones and speech repairs, redefining the speech recognition problem.

Page 16: Presentation1

Advantages (cntd.):A typical NLP system consists of

tokenization, sentence delimitation, part-of-speech (POS) tagging, phrase chunking, parsing, and concept mapping. As one of the initial steps, POS tagging determines the part of speech for each token in a sentence.

Managers, educators, Trainers, Sales people are able to accurately assess the needs of a group, improves questioning techniques thus improving their skills to achieve more consistent results.

Page 17: Presentation1

Limitations :User Cannot enter more than one sentence i.e.

cannot enter paragraph.It is not able to detect and report the gender of

the word i.e. Morphological analysis in not done.

When ambiguity is encountered it is searched for the POS of the ambiguous word if it contains less or no word with the correct POS and there are more number of words for other POS then it shows incorrect POS for the ambiguous word.

Page 18: Presentation1

Applications :Information Retrieval

Speech synthesis

Word Sense Disambiguation (WSD)

Machine Translation (MT)

-Text to Text

-Speech to Speech

Page 19: Presentation1

Snapshots

Page 20: Presentation1
Page 21: Presentation1
Page 22: Presentation1
Page 23: Presentation1

Conclusion and Future Scope :The POS tagger described here is

very simple and efficient for automatic tagging, but the morphological complexity of the Marathi make it hard. The performance of the current system is good and result achieved by this method are excellent. In future we wish to improve the accuracy our system by adding more tagged sentence in our training corpus.