Unsupervised Transfer Learning

So far, we have studied the property of supervised transfer learning for QA, which means that during pre-training and fine-tuning, both the source and target datasets provide the correct answer for each question. We now conduct unsupervised transfer learning experiments described in Section 2.2 (Algorithm 1), where the answers to the questions in the target dataset are not available. We used QACNN as the QA model and all the parameters ($E$, $W_{CNN}^{(1)}$, $W_{CNN}^{(2)}$, $W_{FC}^{(1)}$, and $W_{FC}^{(2)}$) were updated during fine-tuning in this experiment. Since the range of the testing accuracy of the TOFEL-series (TOFEL-manual and TOFEL-ASR) is different from that of MCTest (MC160 and MC500), their results are displayed separately in Figure 3(a) and Figure 3(b), respectively.

Figure 3: The figures show the results of unsupervised transfer learning. The x-axis is the number of training epochs, and the y-axis is the corresponding testing accuracy on the target dataset. When training epoch = 0, the performance of QACNN is equivalent to row (b) in Table 2. The horizontal lines, where each line has the same color to its unsupervised counterpart, are the performances of QACNN with supervised transfer learning (row (e) in Table 2), and are the upper bounds for unsupervised learning.

Experimental Result

From Figure 3(a) and Figure 3(b) we observe that without ground truth in the target dataset for supervised fine-tuning, transfer learning from source dataset can still improve the performance through a simple iterative self-labeling mechanism. For TOFEL-manual and TOFEL-ASR, QACNN achieves the highest testing accuracy at Epoch 7 and 8, outperforming its counterpart without fine-tuning (epoch=0) by approximately 4% and 5%, respectively. For MC160 and MC500, the QACNN achieves the peak at Epoch 3 and 6, outperforming its counterpart without fine-tuning (epoch=0) by about 2% and 6%, respectively. The results also show that the performance of unsupervised transfer learning is still worse than supervised transfer learning, which is not surprising, but the effectiveness of unsupervised transfer learning when no ground truth labels are provided is validated.

Attention Maps Visualization

To better understand the unsupervised transfer learning process of QACNN, we visualize the changes of the word-level attention map during training Epoch 1, 4, 7, and 10 in Figure 4. We use the same question from TOFEL-manual as shown in Table 1 as an example. From Figure 4 we can observe that as the training epochs increase, the QACNN focuses more on the context in the story that is related to the question and the correct answer choice. For example, the correct answer is related to "class project". In epoch 1 and 4, the model does not focus on the phrase "class representation", but the model attends on the phrase in Epoch 7 and 10. This demonstrates that even without ground truth, the iterative process in Algorithm 1 is still able to lead the QA model to gradually focus more on the important part of the story for answering the question.

Figure 4: Visualization of the changes of the word-level attention map in the first stage CNN of QACNN in different training epochs. The more red, the more the QACNN views the word as a key feature. The input story-question-choices triplet is the same as the one in Table 1.