I have about 500 cells of data that have been mapped tro specific themes. Text in cells can range from 3 words to 500 words.
I know I can do topic modeling but I get such crap results. I tried t5 but it really didn’t give much at all and I couldn’t increase the summary length. I tried gpt2 but keep running into a pytorch issue. I tried bert but I always get a server issue. I think both issues may be related to some restraints on my company computer.
Are there any good options that can summarize this many text responses well? Like I could use the gpt4 api but there are security issues so it needs to be somewhat done locally