• No results found

Dual Latent Variable Model for Low Resource Natural Language Generation in Dialogue Systems

N/A
N/A
Protected

Academic year: 2020

Share "Dual Latent Variable Model for Low Resource Natural Language Generation in Dialogue Systems"

Copied!
10
0
0

Loading.... (view fulltext now)

Full text

Loading

Figure

Figure 1: Illustration of proposed variational mod-els as a directed graph. (a) VNLG: joint learn-ing both variational parameters φ and generativemodel parameters θ
Figure 2: The Dual latent variable model consistsof two VAE models: (i) a VNLG (red-dashedbox) is to generate utterances and (ii) a VariationalCNN-DCNN is an auxiliary auto-encoding model(left side)
Figure 3: Performance on Laptop domain with var-ied limited amount, from 1% to 7%, of the adap-tation training data when adapting models pre-trained on [Restaurant+Hotel] union dataset.
Table 1: Results evaluated on four domains by training models from scratch with 10%, 30%, and 100%in-domain data, respectively
+4

References

Related documents

• how they and others choose and adapt features and functions of spoken language for specific purposes in different contexts. • some common influences on spoken language and

[r]

The aim of this study was to evaluate the current vac- cination status of the HCWs in all of the Departments different from the Department for the Health of Women and Children of one

No Australian firms currently have Tokyo offices, nor are they likely to in the near future (although I have heard very unsubstantiated rumours that Minter Ellison is contemplating

Drawing on history, politics, language, literature, contemporary media, music, art, and relationship building with Pacific communities, Pacific Studies and Samoan Studies

Furthermore, while symbolic execution systems often avoid reasoning precisely about symbolic memory accesses (e.g., access- ing a symbolic offset in an array), C OMMUTER ’s test

Adipocytes were preincubated 30 min in the presence (white columns) or absence (black columns) of 1 ␮ mol/l insulin. Values are the means ⫾ sem of 4 separate experiments. B) Effect

Examples (non-statutory)  a depth study linked to one of the British areas of study listed above  a study over time, testing how far sites in their locality reflect aspects