CODEWITHSUNDEEP
javaregexstringalgorithmnlp
devsda
devsda

Reputation: 4222

Interpret sentence and convert into their corresponding format

There are some formats of inputs and their corresponding outputs

1. 7 years 10 months  ---> YRS:7 MNHS:10
2. 7 kgs 10 grms      ---> KGS:7 GRMS:10
3. 7 kilograms 10 grams  ---> KGS:7 GRMS:10
4. 7 thousand 9 hundread ---> 7900
5. seven years ten months --> YRS:7 MNHS:10 
6. seven kgs ten grms     ---> KGS:7 GRMS:10
7. triple seven double five  --> 77755

I wrote separate modules for all by storing informations in **HashMap. And it is working fine.**

Then I need to write one main module in which input is one sentence(utterance), and I need to replace all above substrings into corresponding substring output.

For example,

Input :- Dial number triple eight triple four three nine eight.

Output :- Dial number 888444398.

and many such utterances.

My doubts :-

  1. I used numbers of HashMap for smaller modules to store meaning of keys, just like - triple means 3 times, double means 2 times and all. But this has limitation that if I need to add anything I have to add that entry in HashMap. Suggest some good technique for this.

  2. I am confused in main module, how to extract useful substring given in above examples from given utterances. So suggest some good technique for this also.

Project Lanuguage : Java.

Upvotes: 1

Views: 74

Answers (2)

Daniel
Daniel

Reputation: 6039

You should look at Illinos Quantifier package:

http://cogcomp.cs.illinois.edu/page/software_view/Quantifier

http://cogcomp.cs.illinois.edu/demo/quantities/results.php

Upvotes: 1

Paul
Paul

Reputation: 154

You might want to use some kind of formal grammar parser. Just doing design of a grammar can clear a view of the problem. In the most simple case your grammar could look like:

    STRING -> "" | STRING MEASUREMENT | STRING NUMBER | STRING WORD
    MEASUREMENT -> NUMBER UNITS
    UNITS -> kgs | grms | years | months | ...
    NUMBER -> THOUSAND HUNDRED NUMBER_BELOW_HUNDRED | THOUSAND HUNDRED
    THOUSAND -> "" | NUMBER_BELOW_HUNDRED thousand
    HUNDRED -> "" | NUMBER_BELOW_HUNDRED hundred
    NUMBER_BELOW_HUNDRED -> one | two | three | ... | ninety nine | 99 | 98 | ... | 1
    WORD -> /* all other */

You can write a parser by yourself (in this case it seems to be pretty easy) or use a ready solution like Bison/Flex.

The usual alternative for your HashMaps are configuration files.

Upvotes: 0

Related Questions