How to analyse sentiment according to the aspect term using infer_example.py?

[andrikristian]

HI,

In the file infer_example.py, I have the following code

def evaluate(self, raw_texts):\n ......\n aspect_seqs = [self.tokenizer.text_to_sequence('battery')] * len(raw_texts)\n ......

t_probs = inf.evaluate(['laptop is good but battery is bad'])\n print(t_probs.argmax(axis=-1) - 1)

why is the sentiment score = 1?

I tested the sentence 'laptop is good' and 'battery is bad and the output are 1 and -1 respectively. But when I combine the sentence together, the output is always 1 no matter the aspect term.

The model I am using is AOA, and it is trained on the laptop reviews and the accuracy is 0.7304.

[GeneZC]

Maybe you should give the way you combined them so that we could further give it a guess.

[andrikristian]

infer_example.txt

Attached is the file I used. Thanks.

[GeneZC]

Generally, it's because both 'good' and 'bad' could be descriptive of 'battery', and most models only concentrate on semantic correlations, leading to the result of misidentifying the correct modifier.

[andrikristian]

Thanks. I found out that if I input 'good laptop, bad battery', the sentiment predictions are correct for both aspect terms: 'laptop' and 'battery'.

Is it true that the context words behind the aspect word weigh more for sentiment prediction?

[GeneZC]

Actually, prior works mainly focus on how to obtain a aspect-specific context representation properly so that we could easily judge which part in the context is much more important than the rest.
In this way, from my perspective, aspect and the corresponding context are of equal importance.

[andrikristian]

ok thanks for your answers.