const nlp = wink_nlp_1.default(wink_eng_lite_model_1.default);
^
TypeError: wink_nlp_1.default is not a function
at Object.<anonymous> (/Volumes/DATA/Projects/INhouse/backend/dist/parser/articles/article.parser.js:7:31)
at Module._compile (internal/modules/cjs/loader.js:1063:30)
at Object.Module._extensions..js (internal/modules/cjs/loader.js:1092:10)
at Module.load (internal/modules/cjs/loader.js:928:32)
at Function.Module._load (internal/modules/cjs/loader.js:769:14)
at Module.require (internal/modules/cjs/loader.js:952:19)
at require (internal/modules/cjs/helpers.js:88:18)
Hi there, I think wink is awesome! I wanted to know is there a way to have entity sentiment? for example if there are to subjects in a sentence, instead of doing a "global" sentiment for the whole sentence, do for each entity, for example "London is cold and I don't like it, Paris is warm and nice!" so the entities I assume here is "London" with negetive sentiment and "Paris" with positive sentiment. Any idea how to achieve this? thanks!
Hi @pantchox one possible way could be to (a) split text into sentences, (b) extract sentence wise sentiment score using its.sentiment helper, and finally (c) look for the entity and assign the score to that entity.
@pantchox request you to refer to the winkNLP's getting started documentation.
@codescientist703 You are right, these are 2 projects - a) Writing Assistant and b) Automated Summary & Keyword extraction. In case you need further discussion, please join the slack link for graype queries https://app.slack.com/client/T01V7TZ3NR2/C023X3KUQ2U
Hey everyone, I am Priyaraj, I am interested in working on the project 'winkNLP' as part of the github externship program. I would love some information on what are the things that I should focus on while writing my proposal. How can I get more information related to the project?
Hi @Priyaraj17 Great to hear that! Please join https://app.slack.com/client/T01V7TZ3NR2/C023X3KUQ2U to view our pre-reqs shared on the forum.
Hi all, I was wondering if others have used winkjs with skypack (https://www.skypack.dev/) or unpkg to avoid requiring a bundler in the browser?
I've published an attempt at using it in the browser with an observablehq notebook here: https://observablehq.com/@adamocarolli/wink-nlp It appears like the model has loaded correctly, however, when I use helper functions from
its or as to transform the output I don't get any changes ie. doc.tokens().out( its.type, as.freqTable ) // -> [ 'Hello', 'World', '🌎', '!', 'How', 'are', 'you', '?' ]
@adamocarolli Thanks for your question! Sometime ago we also got stuck and paused further exploration; today with your question, we had an epiphany!! We have released a new version that resolves this problem. The problem was occurring because these helpers are
functions and within the .out() api there were many manipulations on these functions AND on browser these were failing. Now you should not use require to obtain the its and as helpers; instead use winkNLP instance to acquire them. Please see https://observablehq.com/@winkjs/testbed-for-winknlp for exact details.
Excellent! Glad my question was of some help here and thank you for both releasing the new version and providing a test bed @sanjayaksaxena. I have a feeling now that you've enabled winkjs on observable the d3.js community is gonna pick it up and make some amazing data visualization with it!
I had one more question: I'm making the assumption that the model you're using for pos and ner outputs a set of associated confidence scores. Is there a way to retrieve these scores? For example, if winkNLP labels the token
months as a NOUN I'd like to be able to retrieve how confident the model is in that labelling. I've built an example workflow here to demonstrate this idea: https://observablehq.com/@adamocarolli/winknlp-example-confidence-score-interface
let's take the bm25-text-search you implemented here. i want to use this logic but also that it would be able to mark repeated common phrases. and not that i will manually search for it in the text.
Hello @RonDaha, would like you to explore the bm25vectorizer utility. For this you will have to use wink-nlp package instead wink-nlp-utils. The examples on the page are simple to follow and implement.
Hi all. I have been using the wink library built from past couple of days and appreciate the detailing of the functionalities and the ease of use. Just had a question around the wink-naive-bayes-text-classifier. From the example provided it is clear to use and I was able to use it well.
nbc.learn( 'I want to prepay my loan', 'prepay' );
in the above line, can i train it for multiple tokens. like below
nbc.learn( 'I want to prepay my loan', ['prepay', 'loans', 'amount'] );
And also when it is used for predicting, can it return multiple tokens?
nbc.learn( 'I want to prepay my loan', 'prepay' );
in the above line, can i train it for multiple tokens. like below
nbc.learn( 'I want to prepay my loan', ['prepay', 'loans', 'amount'] );
And also when it is used for predicting, can it return multiple tokens?
@niranja33050187_twitter please checkout https://runkit.com/sanjaya/6128e6f8192462001548b330 at RunKit.
1 reply
Hi @stellasphere yes winkNLP has custom entity definition and recognition. Follow this link to get started https://winkjs.org/wink-nlp/custom-entities.html