POS Tagging - Continued

Word-sense-disambiguation

Word-sense-disambiguation (WSD) is a classical NLU problem of deciding in which sense a particular word is used in a sentence. A word can have many senses—for instance, consider the word "bass." Here are some senses we can think of:

• Bass—seabass, fish (noun (N))

• Bass—lowest male voice (N)

• Bass—male singer with lowest voice range (N)

Determining the sense of the word can be crucial in search engines, machine translation, and question-answering systems. For the preceding example, "bass," a POS tagger is unfortunately not of much help as the tagger labels all senses with a noun tag. We need more than a POS tagger. How about the word "beat?" Let's have a look at this here:

• Beat—to strike violently (verb (V))

• Beat—to defeat someone else in a game or a competition (V)

• Beat—rhythm in music or poetry (N)

• Beat—bird wing movement (N)

• Beat—completely exhausted (adjective (ADJ))

Here, POS tagging can help a lot, indeed. The ADJ tag determines the word sense definitely; if the word "beat" is tagged as ADJ, it identifies the "completely exhausted" sense. This is not true for the V and N tags here; if the word "beat" is labeled with a V tag, its sense can be "to strike violently" or "to defeat someone else." WSD is an open problem, and many complicated statistical models are proposed. However, if we need a quick prototype, we can tackle this problem in some cases (such as in the preceding example) with the help of the spaCy tagger.

Verb tense and aspect in NLU applications

In the previous chapter, we used the example of the travel agency application, where we got the base forms (which are freed from verb tense and aspect) of the verbs by using lemmatization. In this section, we'll focus on how to use the verb tense and aspect information that we lost during the lemmatization process.

Verb tense and aspect are maybe the most interesting information that verbs provide us, telling us when the action happened in time and if the action of the verb is finished or ongoing. Tense and aspect together indicate a verb's reference to the current time. English has three basic tenses: past, present, and future. A tense is accompanied by either simple, progressive/continuous, or perfect aspects. For instance, in the sentence, "I'm eating," the action "eat" happens in the present and is ongoing. Hence we describe this verb as present progressive/continuous.

So far, so good. So, how do we use this information in our travel agency NLU, then? Consider the following customer sentences that can be directed to our NLU application: