This week starts a new course on NLP in a popular (and free) on-line learning platform. This is a course I’ve been waiting for years and it could be great.

As I don’t want to do any kind of marketing or spam, I won’t publish the link or the name of the course (anyway, it won’t be hard to google it). If anyone is interested and needs more details, just let me know and it would be great to have a colleague from chatbots in the course!

Regards,

Jose

I think I speak for all when I say this sort of thing isn’t spam. I think it would be of interest to many on the forum. Please feel free to post a link.

I’ll second Steve’s opinion. By all means, please provide any information regarding this that you feel is relevant.

Thanks for your answers! Here is the link:

https://www.coursera.org/learn/natural-language-processing

I think it covers some topics that are really useful for chatbot creation, such as part-of-speech tagging or question answering.

If anyone is interested, I suggest to act quickly, since the closing date for enlisting is soon!

I just signed up, so I look forward to learning alongside you, José!

From the class syllabus:

Week Three: NLP Tasks and Text Similarity
Week Three will cover Vector Semantics, Text Similarity, and Dimensionality Reduction. I will also go through a long list of sample NLP tasks (e.g., Information Extraction, Text Summarization, and Semantic Role Labeling) and introduce each of them briefly.

Reading this, I was thinking:

I want to enter a text, then compare other subsequent texts to it to see whether they are similar enough to warrant the same response.

Count words in each text and report the similarity as number of similar words?

Suppose an array of reference texts linked to potential output texts.

For each word in the new input:

  If that word is in the reference text, add 1 to the similarity total.

  If a word in the input text’s equivalence relations is in the equivalence relations of a word in the reference text, increment the similarity score.

  [Note: equivalence relations can be stored and retrieved in http://subbot.org/logicagent/ .]

  Store the final similarity total for that reference text, go on to the next reference text, and repeat the steps above.

  Choose the reference text with the highest similarity total to generate an output.

—-

Reference texts are trained, or copied and pasted, from the parents of my posts, say. My responses are linked to the reference texts that provoked my response.

Run the similarity algorithm on all the initial reference texts to see if some can be consolidated. (How to “consolidate”?)

—-

Just some thoughts provoked by looking at the syllabus. Does the class provide any useful programs I could plug into my app and play to give me a similarity score between two texts?

I started a new thread in reply to R. Mitchell’s comments.