Research Article
Enhancing Text Generation via Parse Tree Embedding
Table 3
Examples generated by Tree-VAE and baselines on the e-mail corpus dataset.
| | MLE: |
| | sentence1: The company in this month<UNK>its earnings for the past four years. | | sentence2: You might want to check to add value and differentiate in areas other than price to <UNK>going to them also. | | sentence3: Just another update on where we have spent enough time. |
| | SVAE: |
| | sentence1: Phil have been invited to attend he web seminar at the system | | sentence2: Gonzalez will be added to your shopping <UNK>and made the decision back up together. | | sentence3: I expect complete important it by the close of additional databases. |
| | SeqGAN: |
| | sentence1: Follow the instructions to choose a new password | | sentence2: We trust you will enjoy<UNK>our advertised companies. | | sentence3: It works for works on both internal and external company e-mail on how to book this new opportunity. |
| | RankGAN: |
| | sentence1: We currently have two possible offers coming builder due to your strong local. | | sentence2: I will get them signed and send them forward to discussing these meetings. | | sentence3: I am very interested for people to pay the bills as soon as. |
| | Tree-VAE: |
| | sentence1: The followings are the folders that we use in the network drive for all existing computer. | | sentence2: We found this to be the most effective and effective way to deal with tax issues. | | sentence3: Last years, we were experiencing a major move, somany people played a key role in maintaining our team. |
|
|
