NLP (ML-AI)
NLP (natural language processing) – answer NLP consists of developing a set of
algorithms and tools so that machines can make sense of data available in natural
(human) language such as English, German, French, etc.
NLP algorithms often model the hierarchical structure of natural language i.e. characters
form words, words form phrases, phrases form sentences, sentences form paragraphs,
and paragraphs form documents.
Although there are traces of NLP research since a long time ago, the concept got well
defined in the 1950s, with the breakthrough research work of Alan Turing and Noam
Chomsky.
natural language - answer Refers to the medium in which humans communicate with
each other. This could be in the form of writings (text) for example emails, articles,
news, blogs, bank documents, etc or speech for example lectures, speeches, audio
calls, etc.
Common applications of NLP - answer- document classification: classifying textual
documents and assigning it one or multiple categories.
- document clustering: find similar documents and segregate them to form groups.
- sentiment analysis: classify text for different sentiments ranging from negative to
neutral to positive.
- document summarization: extract the most important and central ideas in a document.
- named entity recognition (entity extraction): identifies names entities and classifies
them into categories such as person, organization, location, etc.
- question answering: intelligent systems that generate responses to the questions
being asked by the user.
- machine translation: automatically translating from one natural language to another.
Natural Language Understanding (NLU) - answerNLU enables machines to understand
the intent or the meaning of the text.
This involves the following levels of analysis:
1. Morphological
2. Syntactic
3. Semantic
4. Discourse
Morphological analysis (NLU) - answerIs the analysis of the structure of individual
words.
, A morpheme is defined as the "minimal unit of meaning". For example, the words
"care", "cares", "caring", "careless", "careful", "uncaring" are different forms of the same
word, with stem "care". Also note that the structure of a word could include a prefix
("un-" in "uncaring") or a suffix ("-less" in "careless").
Morpheme - answerA morpheme is defined as the "minimal unit of meaning". For
example, the words "care", "cares", "caring", "careless", "careful", "uncaring" are
different forms of the same word, with stem "care". Also note that the structure of a word
could include a prefix ("un-" in "uncaring") or a suffix ("-less" in "careless").
Syntactic Analysis (NLU) - answer(Also called parsing) involves analysis of words in the
sentence for grammar.
Semantic Analysis (NLU) - answerUses morphological and syntactic knowledge to
understand the meaning, intent, and purpose of the text as a whole.
This is necessary since grammar leaves a lot of ambiguity, and we implicitly rely on
shared understanding of the world to communicate with each other.
For example, consider the two sentences "I went to the market in my shorts" vs "I went
to the market in the city". In terms of grammer (syntax) the two sentences are
equivalent. However, based on semantics (meaning), we know that "in my shorts" refers
to "I", whereas "in the city" refers to the "market".
Discourse Analysis (NLU) - answerIs more advanced stage of NLU where syntactic or
semantic analysis is performed on a longer piece of text.
That is, the analysis is performed over a paragraph or an entire document, as opposed
to a single sentence.
Natural Language Generation (NLG) - answerOnce the machine understands the
natural language, NLG is used to respond in natural language, or to product written text.
In general, NLG systems are more complex than NLU.
Recent applications include chat-bots and personal assistants like Alexa and Siri.
Some general NLG approaches:
- content determination
- planning/micro-planning
- deep learning
Content determination (NLG) - answerInvolves deciding what information we need to
convey in the generated text.
There are pre-built schemas or templates to specify the content. Using knowledge-
based rules and pattern detection, the words in these templates are predicted.
NLP (natural language processing) – answer NLP consists of developing a set of
algorithms and tools so that machines can make sense of data available in natural
(human) language such as English, German, French, etc.
NLP algorithms often model the hierarchical structure of natural language i.e. characters
form words, words form phrases, phrases form sentences, sentences form paragraphs,
and paragraphs form documents.
Although there are traces of NLP research since a long time ago, the concept got well
defined in the 1950s, with the breakthrough research work of Alan Turing and Noam
Chomsky.
natural language - answer Refers to the medium in which humans communicate with
each other. This could be in the form of writings (text) for example emails, articles,
news, blogs, bank documents, etc or speech for example lectures, speeches, audio
calls, etc.
Common applications of NLP - answer- document classification: classifying textual
documents and assigning it one or multiple categories.
- document clustering: find similar documents and segregate them to form groups.
- sentiment analysis: classify text for different sentiments ranging from negative to
neutral to positive.
- document summarization: extract the most important and central ideas in a document.
- named entity recognition (entity extraction): identifies names entities and classifies
them into categories such as person, organization, location, etc.
- question answering: intelligent systems that generate responses to the questions
being asked by the user.
- machine translation: automatically translating from one natural language to another.
Natural Language Understanding (NLU) - answerNLU enables machines to understand
the intent or the meaning of the text.
This involves the following levels of analysis:
1. Morphological
2. Syntactic
3. Semantic
4. Discourse
Morphological analysis (NLU) - answerIs the analysis of the structure of individual
words.
, A morpheme is defined as the "minimal unit of meaning". For example, the words
"care", "cares", "caring", "careless", "careful", "uncaring" are different forms of the same
word, with stem "care". Also note that the structure of a word could include a prefix
("un-" in "uncaring") or a suffix ("-less" in "careless").
Morpheme - answerA morpheme is defined as the "minimal unit of meaning". For
example, the words "care", "cares", "caring", "careless", "careful", "uncaring" are
different forms of the same word, with stem "care". Also note that the structure of a word
could include a prefix ("un-" in "uncaring") or a suffix ("-less" in "careless").
Syntactic Analysis (NLU) - answer(Also called parsing) involves analysis of words in the
sentence for grammar.
Semantic Analysis (NLU) - answerUses morphological and syntactic knowledge to
understand the meaning, intent, and purpose of the text as a whole.
This is necessary since grammar leaves a lot of ambiguity, and we implicitly rely on
shared understanding of the world to communicate with each other.
For example, consider the two sentences "I went to the market in my shorts" vs "I went
to the market in the city". In terms of grammer (syntax) the two sentences are
equivalent. However, based on semantics (meaning), we know that "in my shorts" refers
to "I", whereas "in the city" refers to the "market".
Discourse Analysis (NLU) - answerIs more advanced stage of NLU where syntactic or
semantic analysis is performed on a longer piece of text.
That is, the analysis is performed over a paragraph or an entire document, as opposed
to a single sentence.
Natural Language Generation (NLG) - answerOnce the machine understands the
natural language, NLG is used to respond in natural language, or to product written text.
In general, NLG systems are more complex than NLU.
Recent applications include chat-bots and personal assistants like Alexa and Siri.
Some general NLG approaches:
- content determination
- planning/micro-planning
- deep learning
Content determination (NLG) - answerInvolves deciding what information we need to
convey in the generated text.
There are pre-built schemas or templates to specify the content. Using knowledge-
based rules and pattern detection, the words in these templates are predicted.
