Is it theoretically possible to text mine symptoms following chemical exposure?

Question

[nsnamhan]

Hello world, I am Nam from South Korea (Not VietNam!) and I have a question to ask humbly.

I am conducting research classifying possible body effect following chemical exposure.

I've gone through hundreds of journals to find symptoms following chemical exposure. (e.g. ppm for inhalation and mg for ingestion) So far, I manually conducted the reading and databasing process. However, I recently encountered the possibility of using text mining for my research.

Please kindly provide your opinion on the following matters

1. Is it theoretically possible to text mine symptoms following chemical exposure from pubmed abstract and/or pubmed journal?

2. how can I tokenize sentences instead of words (under the premise that symptoms might be described sentence before or after a sentence containing exposure concentration)

3. what are the general steps to achieve the above explained goals?

4. Are there any material: links, journals, or manual which can possibly answer my question?

Thank you for reading and have a good one!

Answer 1

[Wolf Ihlenfeldt]

My suggestion would be to start your project with safety data sheets instead of full papers or abstracts. They are shorter and more formalized in their content and structure.

Answer 2

[Jean-Karim Heriche]

There's plenty of text mining done in the biomedical domain on all kinds of subjects so it's possible. There's plenty of literature on the subject of text mining as well as online tutorials so do your research. As for tools to get started, check for example Python's NLTK or the R tm package.