# How to generate english sentence from a parse having \*/NN\_u\_unknown,\*/JJ\_u\_unknown?

**URL:** <https://delphinqa.ling.washington.edu/t/how-to-generate-english-sentence-from-a-parse-having-nn-u-unknown-jj-u-unknown/851>\
**Category:** ACE\
**Created:** [November 1, 2022, 1:46pm UTC](https://delphinqa.ling.washington.edu/t/how-to-generate-english-sentence-from-a-parse-having-nn-u-unknown-jj-u-unknown/851 "2022-11-01T13:46:02Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![sv\_Hymavathi](https://delphinqa.ling.washington.edu/letter_avatar_proxy/v4/letter/s/fbc32d/32.png) [@sv\_Hymavathi](https://delphinqa.ling.washington.edu/u/sv_Hymavathi)\
**Post date:** [November 1, 2022, 1:46pm UTC](https://delphinqa.ling.washington.edu/t/how-to-generate-english-sentence-from-a-parse-having-nn-u-unknown-jj-u-unknown/851/1 "2022-11-01T13:46:02Z")

</div>

We are able to generate parse output using ACE for the sentence “I feel a deep sense of gratitude to her.” But when we give the same parse to generate English sentence we get the following error message instead of the expected English sentence:

NOTE: EP ‘\_gratitude/NN\_u\_unknown’ is unknown in the semantic index  
WARNING: EP ‘\_gratitude/NN\_u\_unknown’ is not covered

We are getting same errors for any parse that has RELS like _/NN\_u\_unknown,_/JJ\_u\_unknown.

How can one generate sentences from a parse having english \*/NN\_u\_unknown rels.

Please help.

---

<div class="post-metadata">

**Author:** ![Dan](https://delphinqa.ling.washington.edu/letter_avatar_proxy/v4/letter/d/eada6e/32.png) [@Dan](https://delphinqa.ling.washington.edu/u/Dan)\
**Post date:** [November 1, 2022, 2:33pm UTC](https://delphinqa.ling.washington.edu/t/how-to-generate-english-sentence-from-a-parse-having-nn-u-unknown-jj-u-unknown/851/2 "2022-11-01T14:33:05Z")

</div>

I’m sorry to tell you that there is currently no way to use the ERG to generate from unknown words apart from proper names. For some experiments, you might be able to substitute known words of the same lexical type as the unknown word. Otherwise, you may need to add entries to the ERG’s lexicon, which is possible, but may not be easy to do for some words, since there is not much documentation for this task.

---

<div class="post-metadata">

**Author:** ![guyemerson](https://delphinqa.ling.washington.edu/user_avatar/delphinqa.ling.washington.edu/guyemerson/32/540_2.png) [@guyemerson](https://delphinqa.ling.washington.edu/u/guyemerson)\
**Post date:** [November 1, 2022, 2:59pm UTC](https://delphinqa.ling.washington.edu/t/how-to-generate-english-sentence-from-a-parse-having-nn-u-unknown-jj-u-unknown/851/3 "2022-11-01T14:59:52Z")

</div>

Substituting words (as @Dan suggested) was implemented by Ewa Muszyńska and Matic Horvat, briefly described in their PhD theses:

[https://www.repository.cam.ac.uk/bitstream/handle/1810/312207/thesis.pdf?sequence=2](https://www.repository.cam.ac.uk/bitstream/handle/1810/312207/thesis.pdf?sequence=2) (section 2.4.1)  
[https://www.cl.cam.ac.uk/techreports/UCAM-CL-TR-913.pdf](https://www.cl.cam.ac.uk/techreports/UCAM-CL-TR-913.pdf) (section 7.1.1)