Low accuracy during implementation

[Tushar-Melmalgi]

i wrote a small txt file consisting of the above 5 lines. I used the same processing for creating embeddings and I only passed the text part after the colon. reversed the one hot encodings on the prediction and printed the tag next to the sentences and you can see that its totally inaccurate. The tags each mean -

1. Offers, Options, Commits
2. Maybe/Accept-part
3. Statement-opinion
4. Negative non-no answers
5. Conventional-opening

as you can see this is pretty inaccurate. Can you please help.