FreeTxt: supporting bilingual free-text survey and questionnaire data analysis

在现代消费者主导的文化中，获取和回应定性反馈（通常是自由文本评论/书面反馈）是各行各业的专业实践中不可或缺的一部分。例如，调查被用于工作人员发展、专业培训、产品设计和测试，以及公共和私营部门各种形式的服务提供。调查和问卷经常产生定量和定性数据形式的结合。定量形式，如评分量表（如李克特量表）、选择题和等级顺序问题可以很容易地计算（即量化），分析可以以系统的、通常是自动化的方式进行。相比之下，更多的定性问题，如开放式的、自由文本的评论回复，或者，在旅游和遗产行业的背景下，来自社交媒体渠道或网站（包括Trip Advisor和Trust Pilot）上的展览、活动和/或历史遗址的书面反馈，对分析师构成了更大的挑战。处理书面的、基于文本的反馈通常需要更多的劳动密集型和手工分析方法。使这一挑战更加复杂的是，反馈是用英语和威尔士语提出的，就像威尔士经常出现的情况一样，威尔士代表着英国最大的双语社区。双语数据的成功分析依赖于拥有适当的语言专业知识来处理它的工作人员。虽然有一系列复杂的数字工具可用于分析基于文本的数据，特别是对于在学术界、市场营销和公共关系领域工作的研究人员来说，但所使用的许多数字资源不一定负担得起，也不一定快速易用，并且/或非专业用户也可以访问。具体来说，这些工具目前还不能完全支持系统地处理威尔士语自由文本回复的任务。该项目旨在通过构建新颖的“freetext”工具包来弥合这一差距，该工具包旨在支持多种形式的开放式、自由文本数据的分析和可视化，包括英语和威尔士语。freetext将利用现有的基于开源双语语料库的工具和方法，重新包装这些工具和方法，并将它们带到一个新的方向，以便它们与新的受众/用户群体相关。我们将与项目合作伙伴Cadw和威尔士国民信托基金会密切合作，共同设计、共同建设和测试freetext，以确保该资源符合目的，公平和一致地满足威尔士语和英语反应的需求。我们将利用的现有工具包括作为CorCenCC项目（Corpws Cenedlaethol Cymraeg Cyfoes -当代威尔士国家语料库）一部分开发的工具。这包括CorCenCC的语义（即单个单词和短语的基于意义的分类）和词性（POS -即单个单词和短语的基于语法的分类-例如名词，动词）标记器和威尔士语标记集，以及用于语言查询的语料库功能，等等。这些工具将被整合到一个用户友好的在线界面中，用户可以将其文本粘贴/上传其中，以搜索调查答复和反馈中出现的意义模式；看哪些词在特定的主题、地点、话题中最常用；了解游客对某项服务或景点特别喜欢的地方，以及他们认为可以改进的地方。该工具的最终版本将免费提供，并将根据谁可以使用它和何时使用它进行调整。它将包含通用分析功能，使其能够被任何公共和/或专业公司和机构使用，处理定性调查数据的各种数据集，并将与学术研究人员分析和可视化调查数据相关。该工具的可访问性和可用性将有助于提供实现潜在影响的直接途径。

In a modern consumer-led culture, obtaining and responding to qualitative feedback (i.e. often free-text comments/written feedback) is embedded in the professional practice of many walks of life.Surveys are used, for example, in staff development, professional training, product design and testing, and in various forms of service provision across the public and private sector. Surveys and questionnaires often produce a combination of quantitative and qualitative forms of data. Quantitative forms, such as rating scales (e.g. likert scale responses), multiple choice questions and rank order questions can be numerated (i.e. quantified) with ease, the analysis of which can be conducted in a systematic and often automated way. By contrast, more qualitative questions, which prompt open ended, free-text comment responses, or, in the context of the tourism and heritage sector, written feedback from exhibitions, events and/or historical sites on social media channels or websites including Trip Advisor and Trust Pilot, pose a more difficult challenge for the analyst. Tackling written, text-based feedback often requires a more labour-intensive and manual approach to analysis. Compounding this challenge is where feedback is presented in both English and Welsh, as is often the case in Wales, with Wales representing the largest bilingual community in the UK. The successful analysis of bilingual data relies on the workforce having the appropriate linguistic expertise to process it.While a range of sophisticated digital tools for the analysis of text-based data are available, particularly for researchers working in academia, in marketing and public relations contexts etc., many of the digital resources used are not necessarily affordable, quick and easy to use, and/or accessible to non-expert users. Specifically, these tools currently do not fully support the task of systematically processing free-text responses in Welsh.This project aims to bridge this gap by building the novel 'FreeTxt' toolkit which is designed to support the analysis and visualisation of multiple forms of open-ended, free-text data in both English and Welsh. FreeTxt will draw on existing open-source bilingual corpus-based utilities and methodologies, repackaging these and taking them in a new direction so that they are relevant to new audiences/user-groups. We will work closely with project partners Cadw and National Trust Wales to co-design, co-construct and test FreeTxt to ensure that the resource is fit-for-purpose and fairly and consistently meets the needs of Welsh and English-language responses.Existing tools that we will draw on include those developed as part of the CorCenCC project (Corpws Cenedlaethol Cymraeg Cyfoes - The National Corpus of Contemporary Welsh). This includes CorCenCC's semantic (i.e. meaning based categorisations of individual words and phrases) and part of speech (POS - i.e. grammar-based categorisations of individual words and phrases - e.g. nouns, verbs) taggers and tagsets for Welsh language, and corpus functionalities for the querying of language, amongst others. These tools will be integrated into a user-friendly, online interface that users can paste/upload their texts into, to search for patterns of meaning that emerge in survey responses and feedback; to see which words are most often used in relation to a given theme, place, topic; to understand what visitors particularly enjoyed about a service or attraction, and what they think could be improved.The final version of the tool will be made freely-available and will be adaptable in terms of who can use it and when. It will contain generic analysis features that enable it to be used by any public and/or professional company and institution dealing with varying datasets of qualitative survey data and will be of relevance to academic researchers analysing and visualising survey data. The accessibility and usability of this tool will help provide a direct route to potential impact.