skip to main content


Title: Salience as a Narrative Planning Step Cost Function
Psychological research has demonstrated that as we experience a story several features affect the salience of its events in memory. These features correspond to who? where? when? how? and why? questions about those events. Computational models of salience have been used in interactive narratives to measure which events people most easily remember from the past and which they expect more readily from the future. We use three example domains to show that events in sequences that are solutions to narrative planning problems are generally more salient with each other, and events in non-solution sequences are less salient with each other. This means that measuring the salience of a sequence of actions during planning can serve as an efficient cost function to improve the speed, and perhaps also the quality, of a narrative planner.  more » « less
Award ID(s):
1911053
NSF-PAR ID:
10375772
Author(s) / Creator(s):
;
Date Published:
Journal Name:
Proceedings of the 2022 IEEE Conference on Games
Page Range / eLocation ID:
433 to 440
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    During spatial exploration, neural circuits in the hippocampus store memories of sequences of sensory events encountered in the environment. When sensory information is absent during ‘offline’ resting periods, brief neuronal population bursts can ‘replay’ sequences of activity that resemble bouts of sensory experience. These sequences can occur in either forward or reverse order, and can even include spatial trajectories that have not been experienced, but are consistent with the topology of the environment. The neural circuit mechanisms underlying this variable and flexible sequence generation are unknown. Here we demonstrate in a recurrent spiking network model of hippocampal area CA3 that experimental constraints on network dynamics such as population sparsity, stimulus selectivity, rhythmicity and spike rate adaptation, as well as associative synaptic connectivity, enable additional emergent properties, including variable offline memory replay. In an online stimulus‐driven state, we observed the emergence of neuronal sequences that swept from representations of past to future stimuli on the timescale of the theta rhythm. In an offline state driven only by noise, the network generated both forward and reverse neuronal sequences, and recapitulated the experimental observation that offline memory replay events tend to include salient locations like the site of a reward. These results demonstrate that biological constraints on the dynamics of recurrent neural circuits are sufficient to enable memories of sensory events stored in the strengths of synaptic connections to be flexibly read out during rest and sleep, which is thought to be important for memory consolidation and planning of future behaviour.image

    Key points

    A recurrent spiking network model of hippocampal area CA3 was optimized to recapitulate experimentally observed network dynamics during simulated spatial exploration.

    During simulated offline rest, the network exhibited the emergent property of generating flexible forward, reverse and mixed direction memory replay events.

    Network perturbations and analysis of model diversity and degeneracy identified associative synaptic connectivity and key features of network dynamics as important for offline sequence generation.

    Network simulations demonstrate that population over‐representation of salient positions like the site of reward results in biased memory replay.

     
    more » « less
  2. Intelligent interactive narrative systems coordinate a cast of non-player characters to make the overall story experience meaningful for the player. Narrative generation involves a tradeoff between plot-structure requirements and quality of character behavior, as well as computational efficiency. We study this tradeoff using the example of benchmark problems for narrative planning algorithms. A typical narrative planning problem calls for a sequence of actions that leads to an overall plot goal being met, while also requiring each action to respect constraints that create the appearance of character autonomy. We consider simplified solution definitions that enforce only plot requirements or only character requirements, and we measure how often each of these definitions leads to a solution that happens to meet both types of requirements—i.e., the density with which narrative plans occur among plot- or character-requirement-satisfying sequences. We then investigate whether solution densities can guide the selection of narrative planning algorithms. We compare the performance of two search strategies: one that satisfies plot requirements first and checks character requirements afterward, and one that continuously verifies character requirements. Our results show that comparing solution densities does not by itself predict which of these search strategies will be more efficient in terms of search nodes visited, suggesting that other important factors exist. We discuss what some of these factors could be. Our work opens further investigation into characterizing narrative planning algorithms and how they interact with specific domains. The results also highlight the diversity and difficulty of solving narrative planning problems. 
    more » « less
  3. Abstract—Summarization of long sequences into a concise statement is a core problem in natural language processing, which requires a non-trivial understanding of the weakly structured text. Therefore, integrating crowdsourced multiple users’ comments into a concise summary is even harder because (1) it requires transferring the weakly structured comments to structured knowledge. Besides, (2) the users comments are informal and noisy. In order to capture the long-distance relationships in staggered long sentences, we propose a neural multi-comment summarization (MCS) system that incorporates the sentence relationships via graph heuristics that utilize relation knowledge graphs, i.e., sentence relation graphs (SRG) and approximate discourse graphs (ADG). Motivated by the promising results of gated graph neural networks (GG-NNs) on highly structured data, we develop a GG-NNs with sequence encoder that incorporates SRG or ADG in order to capture the sentence relationships. Specifically, we employ the GG-NNs on both relation knowledge graphs, with the sentence embeddings as the input node features and the graph heuristics as the edges’ weights. Through multiple layerwise propagations, the GG-NNs generate the salience for each sentence from high-level hidden sentence features. Consequently, we use a greedy heuristic to extract salient users’ comments while avoiding the noise in comments. The experimental results show that the proposed MCS improves the summarization performance both quantitatively and qualitatively. 
    more » « less
  4. Human activity around the globe is a growing source of selection pressure on animal behavior and communication systems. Some animals can modify their vocalizations to avoid masking from anthropogenic noise. However, such modifications can also affect the salience of these vocalizations in functional contexts such as competition and mate choice. Such is the case in the well-studied Nuttall's white-crowned sparrow ( Zonotrichia leucophrys nuttalli ), which lives year-round in both urban San Francisco and nearby rural Point Reyes. A performance feature of this species' song is salient in territorial defense, such that higher performance songs elicit stronger responses in simulated territorial intrusions; but songs with lower performance values transmit better in anthropogenic noise. A key question then is whether vocal performance signals male quality and ability to obtain high quality territories in urban populations. We predicted white-crowned sparrows with higher vocal performance will be in better condition and will tend to hold territories with lower noise levels and more species-preferred landscape features. Because white-crowned sparrows are adapted to coastal scrub habitats, we expect high quality territories to contain lower and less dense canopies, less drought, more greenness, and more flat open ground for foraging. To test our predictions, we recorded songs and measured vocal performance and body condition (scaled mass index and fat score) for a set of urban and rural birds ( N = 93), as well as ambient noise levels on their territories. Remote sensing metrics measured landscape features of territories, such as drought stress (NDWI), greenness (NDVI), mean canopy height, maximum height, leaf area density (understory and canopy), slope, and percent bare ground for a 50 m radius on each male territory. We did not find a correlation between body condition and performance but did find a relationship between noise levels and performance. Further, high performers held territories with lower canopies and less dense vegetation, which are species-preferred landscape features. These findings link together fundamental aspects of sexual selection in that habitat quality and the quality of sexually selected signals appear to be associated: males that have the highest performing songs are defending territories of the highest quality. 
    more » « less
  5. Speech processing is highly incremental. It is widely accepted that human listeners continuously use the linguistic context to anticipate upcoming concepts, words, and phonemes. However, previous evidence supports two seemingly contradictory models of how a predictive context is integrated with the bottom-up sensory input: Classic psycholinguistic paradigms suggest a two-stage process, in which acoustic input initially leads to local, context-independent representations, which are then quickly integrated with contextual constraints. This contrasts with the view that the brain constructs a single coherent, unified interpretation of the input, which fully integrates available information across representational hierarchies, and thus uses contextual constraints to modulate even the earliest sensory representations. To distinguish these hypotheses, we tested magnetoencephalography responses to continuous narrative speech for signatures of local and unified predictive models. Results provide evidence that listeners employ both types of models in parallel. Two local context models uniquely predict some part of early neural responses, one based on sublexical phoneme sequences, and one based on the phonemes in the current word alone; at the same time, even early responses to phonemes also reflect a unified model that incorporates sentence-level constraints to predict upcoming phonemes. Neural source localization places the anatomical origins of the different predictive models in nonidentical parts of the superior temporal lobes bilaterally, with the right hemisphere showing a relative preference for more local models. These results suggest that speech processing recruits both local and unified predictive models in parallel, reconciling previous disparate findings. Parallel models might make the perceptual system more robust, facilitate processing of unexpected inputs, and serve a function in language acquisition. MEG Data MEG data is in FIFF format and can be opened with MNE-Python. Data has been directly converted from the acquisition device native format without any preprocessing. Events contained in the data indicate the stimuli in numerical order. Subjects R2650 and R2652 heard stimulus 11b instead of 11. Predictor Variables The original audio files are copyrighted and cannot be shared, but the make_audio folder contains make_clips.py which can be used to extract the exact clips from the commercially available audiobook (ISBN 978-1480555280). The predictors directory contains all the predictors used in the original study as pickled eelbrain objects. They can be loaded in Python with the eelbrain.load.unpickle function. The TextGrids directory contains the TextGrids aligned to the audio files. Source Localization The localization.zip file contains files needed for source localization. Structural brain models used in the published analysis are reconstructed by scaling the FreeSurfer fsaverage brain (distributed with FreeSurfer) based on each subject's `MRI scaling parameters.cfg` file. This can be done using the `mne.scale_mri` function. Each subject's MEG folder contains a `subject-trans.fif` file which contains the coregistration between MEG sensor space and (scaled) MRI space, which is used to compute the forward solution. 
    more » « less